You are on page 1of 247



Warren Siegel
University of Maryland
College Park, Maryland

Present address: State University of New York, Stony Brook˜siegel/plan.html

Preface 6. Classical mechanics
1. Introduction 6.1. Gauge covariant 120
1.1. Motivation 1 6.2. Conformal gauge 122
1.2. Known models (interacting) 3 6.3. Light cone 125
1.3. Aspects 4 Exercises 127
1.4. Outline 6 7. Light-cone quantum mechanics
2. General light cone 7.1. Bosonic 128
2.1. Actions 8 7.2. Spinning 134
2.2. Conformal algebra 10 7.3. Supersymmetric 137
2.3. Poincaré algebra 13 Exercises 145
2.4. Interactions 16 8. BRST quantum mechanics
2.5. Graphs 19 8.1. IGL(1) 146
2.6. Covariantized light cone 20 8.2. OSp(1,1|2) 157
Exercises 23 8.3. Lorentz gauge 160
3. General BRST Exercises 170
3.1. Gauge invariance and 9. Graphs
constraints 25 9.1. External fields 171
3.2. IGL(1) 29 9.2. Trees 177
3.3. OSp(1,1|2) 35 9.3. Loops 190
3.4. From the light cone 38 Exercises 196
3.5. Fermions 45 10. Light-cone field theory 197
3.6. More dimensions 46 Exercises 203
Exercises 51 11. BRST field theory
4. General gauge theories 11.1. Closed strings 204
4.1. OSp(1,1|2) 52 11.2. Components 207
4.2. IGL(1) 62 Exercises 214
4.3. Extra modes 67 12. Gauge-invariant interactions
4.4. Gauge fixing 68 12.1. Introduction 215
4.5. Fermions 75 12.2. Midpoint interaction 217
Exercises 79 Exercises 228
5. Particle References 230
5.1. Bosonic 81 Index 241
5.2. BRST 84
5.3. Spinning 86
5.4. Supersymmetric 95
5.5. SuperBRST 110
Exercises 118

First, I’d like to explain the title of this book. I always hated books whose titles
began “Introduction to...” In particular, when I was a grad student, books titled
“Introduction to Quantum Field Theory” were the most difficult and advanced text-
books available, and I always feared what a quantum field theory book which was
not introductory would look like. There is now a standard reference on relativistic
string theory by Green, Schwarz, and Witten, Superstring Theory [0.1], which con-
sists of two volumes, is over 1,000 pages long, and yet admits to having some major
omissions. Now that I see, from an author’s point of view, how much effort is nec-
essary to produce a non-introductory text, the words “Introduction to” take a more
tranquilizing character. (I have worked on a one-volume, non-introductory text on
another topic, but that was in association with three coauthors.) Furthermore, these
words leave me the option of omitting topics which I don’t understand, or at least
being more heuristic in the areas which I haven’t studied in detail yet.

The rest of the title is “String Field Theory.” This is the newest approach
to string theory, although the older approaches are continuously developing new
twists and improvements. The main alternative approach is the quantum mechanical
(/analog-model/path-integral/interacting-string-picture/Polyakov/conformal- “field-
theory”) one, which necessarily treats a fixed number of fields, corresponding to
homogeneous equations in the field theory. (For example, there is no analog in the
mechanics approach of even the nonabelian gauge transformation of the field theory,
which includes such fundamental concepts as general coordinate invariance.) It is also
an S-matrix approach, and can thus calculate only quantities which are gauge-fixed
(although limited background-field techniques allow the calculation of 1-loop effective
actions with only some coefficients gauge-dependent). In the old S-matrix approach
to field theory, the basic idea was to start with the S-matrix, and then analytically
continue to obtain quantities which are off-shell (and perhaps in more general gauges).
However, in the long run, it turned out to be more practical to work directly with
field theory Lagrangians, even for semiclassical results such as spontaneous symmetry
breaking and instantons, which change the meaning of “on-shell” by redefining the
vacuum to be a state which is not as obvious from looking at the unphysical-vacuum
S-matrix. Of course, S-matrix methods are always valuable for perturbation theory,
but even in perturbation theory it is far more convenient to start with the field theory
in order to determine which vacuum to perturb about, which gauges to use, and what
power-counting rules can be used to determine divergence structure without specific
S-matrix calculations. (More details on this comparison are in the Introduction.)
Unfortunately, string field theory is in a rather primitive state right now, and not
even close to being as well understood as ordinary (particle) field theory. Of course,
this is exactly the reason why the present is the best time to do research in this area.
(Anyone who can honestly say, “I’ll learn it when it’s better understood,” should mark
a date on his calendar for returning to graduate school.) It is therefore simultaneously
the best time for someone to read a book on the topic and the worst time for someone
to write one. I have tried to compensate for this problem somewhat by expanding on
the more introductory parts of the topic. Several of the early chapters are actually
on the topic of general (particle/string) field theory, but explained from a new point
of view resulting from insights gained from string field theory. (A more standard
course on quantum field theory is assumed as a prerequisite.) This includes the use
of a universal method for treating free field theories, which allows the derivation of
a single, simple, free, local, Poincaré-invariant, gauge-invariant action that can be
applied directly to any field. (Previously, only some special cases had been treated,
and each in a different way.) As a result, even though the fact that I have tried to
make this book self-contained with regard to string theory in general means that there
is significant overlap with other treatments, within this overlap the approaches are
sometimes quite different, and perhaps in some ways complementary. (The treatments
of ref. [0.2] are also quite different, but for quite different reasons.)
Exercises are given at the end of each chapter (except the introduction) to guide
the reader to examples which illustrate the ideas in the chapter, and to encourage
him to perform calculations which have been omitted to avoid making the length of
this book diverge.
This work was done at the University of Maryland, with partial support from
the National Science Foundation. It is partly based on courses I gave in the falls of
1985 and 1986. I received valuable comments from Aleksandar Miković, Christian
Preitschopf, Anton van de Ven, and Harold Mark Weiser. I especially thank Barton
Zwiebach, who collaborated with me on most of the work on which this book was
June 16, 1988 Warren Siegel

Originally published 1988 by World Scientific Publishing Co Pte Ltd.
ISBN 9971-50-731-5, 9971-50-731-3 (pbk)
July 11, 2001: liberated, corrected, bookmarks added (to pdf)
1.1. Motivation 1


1.1. Motivation
The experiments which gave us quantum theory and general relativity are now
quite old, but a satisfactory theory which is consistent with both of them has yet
to be found. Although the importance of such a theory is undeniable, the urgency
of finding it may not be so obvious, since the quantum effects of gravity are not
yet accessible to experiment. However, recent progress in the problem has indicated
that the restrictions imposed by quantum mechanics on a field theory of gravitation
are so stringent as to require that it also be a unified theory of all interactions, and
thus quantum gravity would lead to predictions for other interactions which can be
subjected to present-day experiment. Such indications were given by supergravity
theories [1.1], where finiteness was found at some higher-order loops as a consequence
of supersymmetry, which requires the presence of matter fields whose quantum effects
cancel the ultraviolet divergences of the graviton field. Thus, quantum consistency led
to higher symmetry which in turn led to unification. However, even this symmetry was
found insufficient to guarantee finiteness at all loops [1.2] (unless perhaps the graviton
were found to be a bound-state of a truly finite theory). Interest then returned to
theories which had already presented the possibility of consistent quantum gravity
theories as a consequence of even larger (hidden) symmetries: theories of relativistic
strings [1.3-5]. Strings thus offer a possibility of consistently describing all of nature.
However, even if strings eventually turn out to disagree with nature, or to be too
intractable to be useful for phenomenological applications, they are still the only
consistent toy models of quantum gravity (especially for the theory of the graviton
as a bound state), so their study will still be useful for discovering new properties of
quantum gravity.
The fundamental difference between a particle and a string is that a particle is a 0-
dimensional object in space, with a 1-dimensional world-line describing its trajectory
in spacetime, while a string is a (finite, open or closed) 1-dimensional object in space,
which sweeps out a 2-dimensional world-sheet as it propagates through spacetime:

x x(τ ) #
particle r #

X(σ) X(σ, τ ) #
# ##
# #
string c


The nontrivial topology of the coordinates describes interactions. A string can be
either open or closed, depending on whether it has 2 free ends (its boundary) or is
a continuous ring (no boundary), respectively. The corresponding spacetime figure
is then either a sheet or a tube (and their combinations, and topologically more
complicated structures, when they interact).
Strings were originally intended to describe hadrons directly, since the observed
spectrum and high-energy behavior of hadrons (linearly rising Regge trajectories,
which in a perturbative framework implies the property of hadronic duality) seems
realizable only in a string framework. After a quark structure for hadrons became
generally accepted, it was shown that confinement would naturally lead to a string
formulation of hadrons, since the topological expansion which follows from using
1/N color as a perturbation parameter (the only dimensionless one in massless QCD,
besides 1/N f lavor ), after summation in the other parameter (the gluon coupling, which
becomes the hadronic mass scale after dimensional transmutation), is the same per-
1.2. Known models (interacting) 3

turbation expansion as occurs in theories of fundamental strings [1.6]. Certain string
theories can thus be considered alternative and equivalent formulations of QCD, just
as general field theories can be equivalently formulated either in terms of “funda-
mental” particles or in terms of the particles which arise as bound states. However,
in practice certain criteria, in particular renormalizability, can be simply formulated
only in one formalism: For example, QCD is easier to use than a theory where gluons
are treated as bound states of self-interacting quarks, the latter being a nonrenor-
malizable theory which needs an unwieldy criterion (“asymptotic safety” [1.7]) to
restrict the available infinite number of couplings to a finite subset. On the other
hand, atomic physics is easier to use as a theory of electrons, nuclei, and photons
than a formulation in terms of fields describing self-interacting atoms whose exci-
tations lie on Regge trajectories (particularly since QED is not confining). Thus,
the choice of formulation is dependent on the dynamics of the particular theory, and
perhaps even on the region in momentum space for that particular application: per-
haps quarks for large transverse momenta and strings for small. In particular, the
running of the gluon coupling may lead to nonrenormalizability problems for small
transverse momenta [1.8] (where an infinite number of arbitrary couplings may show
up as nonperturbative vacuum values of operators of arbitrarily high dimension), and
thus QCD may be best considered as an effective theory at large transverse momenta
(in the same way as a perturbatively nonrenormalizable theory at low energies, like
the Fermi theory of weak interactions, unless asymptotic safety is applied). Hence, a
string formulation, where mesons are the fundamental fields (and baryons appear as
skyrmeon-type solitons [1.9]) may be unavoidable. Thus, strings may be important
for hadronic physics as well as for gravity and unified theories; however, the presently
known string models seem to apply only to the latter, since they contain massless
particles and have (maximum) spacetime dimension D = 10 (whereas confinement in
QCD occurs for D ≤ 4).

1.2. Known models (interacting)

Although many string theories have been invented which are consistent at the
tree level, most have problems at the one-loop level. (There are also theories which
are already so complicated at the free level that the interacting theories have been
too difficult to formulate to test at the one-loop level, and these will not be discussed
here.) These one-loop problems generally show up as anomalies. It turns out that
the anomaly-free theories are exactly the ones which are finite. Generally, topologi-

cal arguments based on reparametrization invariance (the “stretchiness” of the string
world sheet) show that any multiloop string graph can be represented as a tree graph
with many one-loop insertions [1.10], so all divergences should be representable as just
one-loop divergences. The fact that one-loop divergences should generate overlapping
divergences then implies that one-loop divergences cause anomalies in reparametriza-
tion invariance, since the resultant multi-loop divergences are in conflict with the
one-loop-insertion structure implied by the invariance. Therefore, finiteness should
be a necessary requirement for string theories (even purely bosonic ones) in order to
avoid anomalies in reparametrization invariance. Furthermore, the absence of anoma-
lies in such global transformations determines the dimension of spacetime, which in
all known nonanomalous theories is D = 10. (This is also known as the “critical,” or
maximum, dimension, since some of the dimensions can be compactified or otherwise
made unobservable, although the number of degrees of freedom is unchanged.)

In fact, there are only four such theories:
I: N=1 supersymmetry, SO(32) gauge group, open [1.11]
IIA,B: N=2 nonchiral or chiral supersymmetry [1.12]
heterotic: N=1 supersymmetry, SO(32) or E8 ⊗E8 [1.13]
or broken N=1 supersymmetry, SO(16)⊗SO(16) [1.14]

All except the first describe only closed strings; the first describes open strings, which
produce closed strings as bound states. (There are also many cases of each of these
theories due to the various possibilities for compactification of the extra dimensions
onto tori or other manifolds, including some which have tachyons.) However, for sim-
plicity we will first consider certain inconsistent theories: the bosonic string, which has
global reparametrization anomalies unless D = 26 (and for which the local anomalies
described above even for D = 26 have not yet been explicitly derived), and the spin-
ning string, which is nonanomalous only when it is truncated to the above strings.
The heterotic strings are actually closed strings for which modes propagating in the
clockwise direction are nonsupersymmetric and 26-dimensional, while the counter-
clockwise ones are N = 1 (perhaps-broken) supersymmetric and 10-dimensional, or
vice versa.

1.3. Aspects

There are several aspects of, or approaches to, string theory which can best be
classified by the spacetime dimension in which they work: D = 2, 4, 6, 10. The 2D
1.3. Aspects 5

approach is the method of first-quantization in the two-dimensional world sheet swept
out by the string as it propagates, and is applicable solely to (second-quantized) per-
turbation theory, for which it is the only tractable method of calculation. Since it
discusses only the properties of individual graphs, it can’t discuss properties which
involve an unfixed number of string fields: gauge transformations, spontaneous sym-
metry breaking, semiclassical solutions to the string field equations, etc. Also, it can
describe only the gauge-fixed theory, and only in a limited set of gauges. (However,
by introducing external particle fields, a limited amount of information on the gauge-
invariant theory can be obtained.) Recently most of the effort in this area has been
concentrated on applying this approach to higher loops. However, in particle field
theory, particularly for Yang-Mills, gravity, and supersymmetric theories (all of which
are contained in various string theories), significant (and sometimes indispensable)
improvements in higher-loop calculations have required techniques using the gauge-
invariant field theory action. Since such techniques, whose string versions have not
yet been derived, could drastically affect the S-matrix techniques of the 2D approach,
we do not give the most recent details of the 2D approach here, but some of the basic
ideas, and the ones we suspect most likely to survive future reformulations, will be
described in chapters 6-9.

The 4D approach is concerned with the phenomenological applications of the
low-energy effective theories obtained from the string theory. Since these theories are
still very tentative (and still too ambiguous for many applications), they will not be
discussed here. (See [1.15,0.1].)

The 6D approach describes the compactifications (or equivalent eliminations) of
the 6 additional dimensions which must shrink from sight in order to obtain the
observed dimensionality of the macroscopic world. Unfortunately, this approach has
several problems which inhibit a useful treatment in a book: (1) So far, no justification
has been given as to why the compactification occurs to the desired models, or to
4 dimensions, or at all; (2) the style of compactification (Kaluża-Klein, Calabi-Yau,
toroidal, orbifold, fermionization, etc.) deemed most promising changes from year
to year; and (3) the string model chosen to compactify (see previous section) also
changes every few years. Therefore, the 6D approach won’t be discussed here, either
(see [1.16,0.1]).

What is discussed here is primarily the 10D approach, or second quantization,
which seeks to obtain a more systematic understanding of string theory that would
allow treatment of nonperturbative as well as perturbative aspects, and describe the

enlarged hidden gauge symmetries which give string theories their finiteness and other
unusual properties. In particular, it would be desirable to have a formalism in which
all the symmetries (gauge, Lorentz, spacetime supersymmetry) are manifest, finiteness
follows from simple power-counting rules, and all possible models (including possible
4D models whose existence is implied by the 1/N expansion of QCD and hadronic
duality) can be straightforwardly classified. In ordinary (particle) supersymmetric
field theories [1.17], such a formalism (superfields or superspace) has resulted in much
simpler rules for constructing general actions, calculating quantum corrections (su-
pergraphs), and explaining all finiteness properties (independent from, but verified by,
explicit supergraph calculations). The finiteness results make use of the background
field gauge, which can be defined only in a field theory formulation where all symme-
tries are manifest, and in this gauge divergence cancellations are automatic, requiring
no explicit evaluation of integrals.

1.4. Outline

String theory can be considered a particular kind of particle theory, in that its
modes of excitation correspond to different particles. All these particles, which differ
in spin and other quantum numbers, are related by a symmetry which reflects the
properties of the string. As discussed above, quantum field theory is the most com-
plete framework within which to study the properties of particles. Not only is this
framework not yet well understood for strings, but the study of string field theory has
brought attention to aspects which are not well understood even for general types of
particles. (This is another respect in which the study of strings resembles the study
of supersymmetry.) We therefore devote chapts. 2-4 to a general study of field theory.
Rather than trying to describe strings in the language of old quantum field theory,
we recast the formalism of field theory in a mold prescribed by techniques learned
from the study of strings. This language clarifies the relationship between physical
states and gauge degrees of freedom, as well as giving a general and straightforward
method for writing free actions for arbitrary theories.
In chapts. 5-6 we discuss the mechanics of the particle and string. As mentioned
above, this approach is a useful calculational tool for evaluating graphs in perturba-
tion theory, including the interaction vertices themselves. The quantum mechanics
of the string is developed in chapts. 7-8, but it is primarily discussed directly as an
operator algebra for the field theory, although it follows from quantization of the clas-
sical mechanics of the previous chapter, and vice versa. In general, the procedure of
1.4. Outline 7

first-quantization of a relativistic system serves only to identify its constraint algebra,
which directly corresponds to both the field equations and gauge transformations of
the free field theory. However, as described in chapts. 2-4, such a first-quantization
procedure does not exist for general particle theories, but the constraint system can
be derived by other means. The free gauge-covariant theory then follows in a straight-
forward way. String perturbation theory is discussed in chapt. 9.

Finally, the methods of chapts. 2-4 are applied to strings in chapts. 10-12, where
string field theory is discussed. These chapters are still rather introductory, since
many problems still remain in formulating interacting string field theory, even in the
light-cone formalism. However, a more complete understanding of the extension of the
methods of chapts. 2-4 to just particle field theory should help in the understanding
of strings.

Chapts. 2-5 can be considered almost as an independent book, an attempt at a
general approach to all of field theory. For those few high energy physicists who are
not intensely interested in strings (or do not have high enough energy to study them),
it can be read as a new introduction to ordinary field theory, although familiarity with
quantum field theory as it is usually taught is assumed. Strings can then be left for
later as an example. On the other hand, for those who want just a brief introduction
to strings, a straightforward, though less elegant, treatment can be found via the
light cone in chapts. 6,7,9,10 (with perhaps some help from sects. 2.1 and 2.5). These
chapters overlap with most other treatments of string theory. The remainder of the
book (chapts. 8,11,12) is basically the synthesis of these two topics.


2.1. Actions

Before discussing the string we first consider some general properties of gauge
theories and field theories, starting with the light-cone formalism.
In general, light-cone field theory [2.1] looks like nonrelativistic field theory. Using
light-cone notation, for vector indices a and the Minkowski inner product A · B =
η ab Ab B a = Aa B a ,

a = (+, −, i) , A · B = A+ B − + A− B + + Ai B i , (2.1.1)

we interpret x+ as being the “time” coordinate (even though it points in a lightlike
direction), in terms of which the evolution of the system is described. The metric
can be diagonalized by A± ≡ 2−1/2 (A1 ∓ A0 ). For positive energy E(= p0 = −p0 ),
we have on shell p+ ≥ 0 and p− ≤ 0 (corresponding to paths with ∆x+ ≥ 0 and
∆x− ≤ 0), with the opposite signs for negative energy (antiparticles). For example,
for a real scalar field the lagrangian is rewritten as
pi 2 + m2
− 12 φ(p2 2
+ m )φ = −φp+ p− + φ = −φp+ (p− + H)φ , (2.1.2)
where the momentum pa ≡ i∂ a , p− = i∂/∂x+ with respect to the “time” x+ , and
p+ appears like a mass in the “hamiltonian” H. (In the light-cone formalism, p+
is assumed to be invertible.) Thus, the field equations are first-order in these time
derivatives, and the field satisfies a nonrelativistic-style Schrödinger equation. The
field equation can then be solved explicitly: In the free theory,

φ(x+ ) = eix+ H φ(0) . (2.1.3)

p− can then be effectively replaced with −H. Note that, unlike the nonrelativistic
case, the hamiltonian H, although hermitian, is imaginary (in coordinate space), due
to the i in p+ = i∂ + . Thus, (2.1.3) is consistent with a (coordinate-space) reality
condition on the field.
2.1. Actions 9

For a spinor, half the components are auxiliary (nonpropagating, since the field
equation is only first-order in momenta), and all auxiliary components are eliminated
in the light-cone formalism by their equations of motion (which, by definition, don’t
involve inverting time derivatives p− ):
1 1/4
σ i pi + im 1/4
− 12 ψ̄(/
p+ im)ψ = − 22 ( ψ+ †
ψ− †
) √ 2
σ i pi − im − 2p+ ψ−
† †
= − ψ + p− ψ + + ψ − p+ ψ −
1 1
− √ ψ − † (σ i pi − im)ψ + − √ ψ + † (σ i pi + im)ψ −
2 2
→ − ψ + † (p− + H)ψ + , (2.1.4)

where H is the same hamiltonian as in (2.1.2). (There is an extra overall factor of 2
in (2.1.4) for complex spinors. We have assumed real (Majorana) spinors.)
For the case of Yang-Mills, the covariant action is
1  D
S = 2 d x tr L , L = 14 F ab 2 , (2.1.5a)
F ab ≡ [∇a , ∇b ] , ∇a ≡ pa + Aa , ∇a  = eiλ ∇a e−iλ . (2.1.5b)
(Contraction with a matrix representation of the group generators is implicit.) The
light-cone gauge is then defined as

A+ = 0 . (2.1.6)

Since the gauge transformation of the gauge condition doesn’t involve the time deriva-
tive ∂ − , the Faddeev-Popov ghosts are nonpropagating, and can be ignored. The field
equation of A− contains no time derivatives, so A− is an auxiliary field. We therefore
eliminate it by its equation of motion:
0 = [∇a , F +a ] = p+ 2 A− + [∇i , p+ Ai ] → A− = − [∇i , p+ Ai ] . (2.1.7)
p+ 2
The only remaining fields are Ai , corresponding to the physical transverse polariza-
tions. The lagrangian is then

L = 12 Ai 2 Ai + [Ai , Aj ]pi Aj + 14 [Ai , Aj ]2
1 1 1
+ (pj Aj ) [Ai , p+ Ai ] + 2 [Ai , p+ Ai ] . (2.1.8)
p+ p+

In fact, for arbitrary spin, after gauge-fixing (A+··· = 0) and eliminating auxiliary
fields (A−··· = · · ·), we get for the free theory

L = −ψ † (p+ )k (p− + H)ψ , (2.1.9)

where k = 1 for bosons and 0 for fermions.
The choice of light-cone gauges in particle mechanics will be discussed in chapt. 5,
and for string mechanics in sect. 6.3 and chapt. 7. Light-cone field theory for strings
will be discussed in chapt. 10.

2.2. Conformal algebra
Since the free kinetic operator of any light-cone field is just 2 (up to factors of
∂ + ), the only nontrivial part of any free light-cone field theory is the representation
of the Poincaré group ISO(D−1,1) (see, e.g., [2.2]). In the next section we will
derive this representation for arbitrary massless theories (and will later extend it
to the massive case) [2.3]. These representations are nonlinear in the coordinates,
and are constructed from all the irreducible (matrix) representations of the light-
cone’s SO(D−2) rotation subgroup of the spin part of the SO(D−1,1) Lorentz group.
One simple method of derivation involves the use of the conformal group, which is
SO(D,2) for D-dimensional spacetime (for D > 2). We therefore use SO(D,2) notation
by writing (D+2)-dimensional vector indices which take the values ± as well as the
usual D a’s: A = (±, a). The metric is as in (2.1.1) for the ± indices. (These ±’s
should not be confused with the light-cone indices ±, which are related but are a
subset of the a’s.) We then write the conformal group generators as

J AB = (J +a = −ipa , J −a = −iK a , J −+ = ∆, J ab ) , (2.2.1)

where J ab are the Lorentz generators, ∆ is the dilatation generator, and K a are
the conformal boosts. An obvious linear coordinate representation in terms of D+2
coordinates is
J AB = x[A ∂ B] + M AB , (2.2.2)
where [ ] means antisymmetrization and M AB is the intrinsic (matrix, or coordinate-
independent) part (with the same commutation relations that follow directly for the
orbital part). The usual representation in terms of D coordinates is obtained by
imposing the SO(D,2)-covariant constraints

xA xA = xA ∂ A = M A B xB + dxA = 0 (2.2.3a)

for some constant d (the canonical dimension, or scale weight). Corresponding to
these constraints, which can be solved for everything with a “−” index, are the
“gauge conditions” which determine everything with a “+” index but no “−” index:

∂ + = x+ − 1 = M +a = 0 . (2.2.3b)
2.2. Conformal algebra 11

This gauge can be obtained by a unitary transformation. The solution to (2.2.3) is
J +a = ∂ a , J −a = − 12 xb 2 ∂ a + xa xb ∂ b + M a b xb + dxa ,

J −+ = xa ∂ a + d , J ab = x[a ∂ b] + M ab . (2.2.4)

This realization can also be obtained by the usual coset space methods (see, e.g.,
[2.4]), for the space SO(D,2)/ISO(D-1,1)⊗GL(1). The subgroup corresponds to all the
generators except J +a . One way to perform this construction is: First assign the coset
space generators J +a to be partial derivatives ∂ a (since they all commute, according
to the commutation relations which follow from (2.2.2)). We next equate this first-
quantized coordinate representation with a second-quantized field representation: In
0 = δ xΦ = JxΦ + xJΦ
→ J xΦ = JxΦ = −Jˆ xΦ = − xJˆΦ , (2.2.5)

where J (which acts directly on x|) is expressed in terms of the coordinates and their
derivatives (plus “spin” pieces), while Jˆ (which acts directly on |Φ
) is expressed in
terms of the fields Φ and their functional derivatives. The minus sign expresses the
usual relation between active and passive transformations. The structure constants
of the second-quantized algebra have the same sign as the first-quantized ones. We
can then solve the “constraint” J +a = −Jˆ+a on x|Φ
 a Jˆ
xΦ ≡ Φ(x) = UΦ(0) = e−x +a
Φ(0) . (2.2.6)

The other generators can then be determined by evaluating

JΦ(x) = −JˆΦ(x) → U −1 JUΦ(0) = −U −1 JUΦ(0)
ˆ . (2.2.7)

On the left-hand side, the unitary transformation replaces any ∂ a with a −Jˆ+a (the
∂ a itself getting killed by the Φ(0)). On the right-hand side, the transformation gives
terms with x dependence and other J’s ˆ (as determined by the commutator algebra).
(The calculations are performed by expressing the transformation as a sum of multiple
commutators, which in this case has a finite number of terms.) The net result is
(2.2.4), where d is −Jˆ−+ on Φ(0), M ab is −Jˆab , and J −a can have the additional term
−Jˆ−a . However, Jˆ−a on Φ(0) can be set to zero consistently in (2.2.4), and does
vanish for physically interesting representations.

From now on, we use ± as in the light-cone notation, not SO(D,2) notation.

The conformal equations of motion are all those which can be obtained from
pa = 0 by conformal transformations (or, equivalently, the irreducible tensor op-
erator quadratic in conformal generators which includes p2 as a component). Since
conformal theories are a subset of massless ones, the massless equations of motion are
a subset of the conformal ones (i.e., the massless theories satisfy fewer constraints).
In particular, since massless theories are scale invariant but not always invariant un-
der conformal boosts, the equations which contain the generators of conformal boosts
must be dropped.
The complete set of equations of motion for an arbitrary massless representation
of the Poincaré group are thus obtained simply by performing a conformal boost on
the defining equation, p2 = 0 [2.5,6]:
0 = 12 [K a , p2 ] = 12 {J a b , pb } + 12 {∆, pa } = M a b pb + d − pa . (2.2.8)
d is determined by the requirement that the representation be nontrivial (for other
values of d this equation implies p = 0). For nonzero spin (M ab = 0) this equation
implies p2 = 0 by itself. For example, for scalars the equation implies only d =
(D − 2)/2. For a Dirac spinor, M ab = 14 [γ a , γ b ] implies d = (D − 1)/2 and the Dirac
equation (in the form γ a γ · pψ = 0). For a second-rank antisymmetric tensor, we
find d = D/2 and Maxwell’s equations. In this covariant approach to solving these
equations, all the solutions are in terms of field strengths, not gauge fields (since the
latter are not unitary representations). We can solve these equations in light-cone
notation: Choosing a reference frame where the only nonvanishing component of the
momentum is p+ , (2.2.8) reduces to the equations M −i = 0 and M −+ = d−(D −2)/2.
The equation M −i = 0 says that the only nonvanishing components are the ones with
as many (lower) “+” indices as possible (and for spinors, project with γ + ), and no
“−” indices. In terms of Young tableaux, this means 1 “+” for each column. M −+
then just counts the number of “+” ’s (plus 1/2 for a γ + -projected spinor index), so
we find that d − (D − 2)/2 = the number of columns (+ 1/2 for a spinor). We also
find that the on-shell gauge field is the representation found by subtracting one box
from each column of the Young tableau, and in the field strength those subtracted
indices are associated with factors of momentum.
These results for massless representations can be extended to massive represen-
tations by the standard trick of adding one spatial dimension and constraining the
extra momentum component to be the mass (operator): Writing

a → (a, m) , pm = M , (2.2.9)
2.3. Poincaré algebra 13

where the index m takes one value, p2 = 0 becomes p2 + M 2 = 0, and (2.2.8) becomes
M a pb + M am M + d −
pa = 0 . (2.2.10)
The fields (or states) are now representations of an SO(D,1) spin group generated
by M ab and M am (instead of the usual SO(D-1,1) of just M ab for the massless case).
The fields additional to those obtained in the massless case (on-shell field strengths)
correspond to the on-shell gauge fields in the massless limit, resulting in a first-order
formalism. For example, for spin 1 the additional field is the usual vector. For spin
2, the extra fields correspond to the on-shell, and thus traceless, parts of the Lorentz
connection and metric tensor.

For field theory, we’ll be interested in real representations. For the massive case,
since (2.2.9) forces us to work in momentum space with respect to pm , the reality
condition should include an extra factor of the reflection operator which reverses the
“m” direction. For example, for tensor fields, those components with an odd number
of m indices should be imaginary (and those with an even number real).

In chapt. 4 we’ll show how to obtain the off-shell fields, and thus the trace parts,
by working directly in terms of the gauge fields. The method is based on the light-cone
representation of the Poincaré algebra discussed in the next section.

2.3. Poincaré algebra

In contrast to the above covariant approach to solving (2.2.8,10), we now consider
solving them in unitary gauges (such as the light-cone gauge), since in such gauges
the gauge fields are essentially field strengths anyway because the gauge has been
fixed: e.g., for Yang-Mills Aa = ∇+ −1 F +a , since A+ = 0. In such gauges we work
in terms of only the physical degrees of freedom (as in the case of the on-shell field
strengths), which satisfy p2 = 0 (unlike the auxiliary degrees of freedom, which satisfy
algebraic equations, and the gauge degrees of freedom, which don’t appear in any field

In the light-cone formalism, the object is to construct all the Poincaré generators
from just the manifest ones of the (D − 2)-dimensional Poincaré subgroup, p+ , and
the coordinates conjugate to these momenta. The light-cone gauge is imposed by the
M +i = 0 , (2.3.1)

which, when acting on the independent fields (those with only i indices), says that
all fields with + indices have been set to vanish. The fields with − indices (auxiliary
fields) are then determined as usual by the field equations: by solving (2.2.8) for M −i .
The solution to the i, +, and − parts of (2.2.8) gives
M −i = (M i j pj + kpi ) ,

M −+ = d − ≡k ,
kp2 = 0 . (2.3.2)

If (2.2.8) is solved without the condition (2.3.1), then M +i can still be removed (and
(2.3.2) regained) by a unitary transformation. (In a first-quantized formalism, this
corresponds to a gauge choice: see sect. 5.3 for spin 1/2.) The appearance of k is
related to ordering ambiguities, and we can also choose M −+ = 0 by a nonunitary
transformation (a rescaling of the field by a power of p+ ). Of course, we also solve
p2 = 0 as
pi 2
p− = − . (2.3.3)
These equations, together with the gauge condition for M +i , determine all the Poincaré
generators in terms of M ij , pi , p+ , xi , and x− . In the orbital pieces of J ab , x+ can be
set to vanish, since p− is no longer conjugate: i.e., we work at “time” x+ = 0 for the
“hamiltonian” p− , or equivalently in the Schrödinger picture. (Of course, this also
corresponds to removing x+ by a unitary transformation, i.e., a time translation via
p− . This is also a gauge choice in a first-quantized formalism: see sect. 5.1.) The
final result is
pi 2
pi = i∂ i , p+ = i∂ + , p− = − ,
J ij = −ix[i pj] + M ij , J +i = ixi p+ , J −+ = −ix− p+ + k ,
pj 2 1
J −i = −ix− pi − ixi + (M i j pj + kpi ) . (2.3.4)
2p+ p+
The generators are (anti)hermitian for the choice k = 12 ; otherwise, the Hilbert space
metric must include a factor of p+ 1−2k , with respect to which all the generators are
pseudo(anti)hermitian. In this light-cone approach to Poincaré representations, where
we work with the fundamental fields rather than field strengths, k = 0 for bosons and
2 for fermions (giving the usual dimensions d = 12 (D − 2) for bosons and 12 (D − 1) for
fermions), and thus the metric is p+ for bosons and 1 for fermions, so the light-cone
kinetic operator (metric)·2(i∂ − − p− ) ∼ 2 for bosons and 2 /p+ for fermions.
2.3. Poincaré algebra 15

This construction of the D-dimensional Poincaré algebra in terms of D−1 coor-
dinates is analogous to the construction in the previous section of the D-dimensional
conformal algebra SO(D,2) in terms of D coordinates, except that in the conformal
case (1) we start with D+2 coordinates instead of D, (2) x’s and p’s are switched,
and (3) the further constraint x · p = 0 and gauge condition x+ = 1 are used. Thus,
J ab of (2.3.4) becomes J AB of (2.2.4) if x− is replaced with −(1/p+ )xj pj , p+ is set
to 1, and we then switch p → x, x → −p. Just as the conformal representation
(2.2.4) can be obtained from the Poincaré representation (in 2 extra dimensions, by
i → a) (2.3.4) by eliminating one coordinate (x− ), (2.3.4) can be reobtained from
(2.2.4) by reintroducing this coordinate: First choose d = −ix− p+ + k. Then switch
xi → pi , pi → −xi . Finally, make the (almost unitary) transformation generated by
exp[−ipi xi (ln p+ )], which takes xi → p+ xi , pi → pi /p+ , x− → x− + pi xi /p+ .

To extend these results to arbitrary representations, we use the trick (2.2.9), or
directly solve (2.2.10), giving the light-cone form of the Poincaré algebra for arbitrary
representations: (2.3.4) becomes

pi 2 + M 2
pi = i∂ i , p+ = i∂ + , p− = − ,

J ij = −ix[i pj] + M ij , J +i = ixi p+ , J −+ = −ix− p+ + k ,

pj 2 + M 2 1
J −i = −ix− pi − ixi + (M i j pj + M im M + kpi ) . (2.3.5)
2p+ p+
Thus, massless irreducible representations of the Poincaré group ISO(D−1,1) are ir-
reducible representations of the spin subgroup SO(D−2) (generated by M ij ) which
also depend on the coordinates (xi , x− ), and irreducible massive ones are irreducible
representations of the spin subgroup SO(D−1) (generated by (M ij , M im )) for some
nonvanishing constant M. Notice that the introduction of masses has modified only
p− and J −i . These are also the only generators modified when interactions are intro-
duced, where they become nonlinear in the fields.

The light-cone representation of the Poincaré algebra will be used in sect. 3.4
to derive BRST algebras, used for enforcing unitarity in covariant formalisms, which
in turn will be used extensively to derive gauge-invariant actions for particles and
strings in the following chapters. The general light-cone analysis of this section will
be applied to the special case of the free string in chapt. 7.

2.4. Interactions

For interacting theories, the derivation of the Poincaré algebra is not so general,
but depends on the details of the particular type of interactions in the theory. We
again consider the case of Yang-Mills. Since only p− and J −i obtain interacting
contributions, we consider the derivation of only those operators. The expression for
p− Ai is then given directly by the field equation of Ai

0 = [∇a , F ai ] = [∇j , F ji] + [∇+ , F −i] + [∇− , F +i] = [∇j , F ji] + 2[∇+ , F −i] + [∇i , F +− ]

1 j

→ p− Ai = [∇i , A− ] − [∇ , F ji] + [∇i , p+ A− ] , (2.4.1)
where we have used the Bianchi identity [∇[+ , F −i] ] = 0. This expression for p− is
also used in the orbital piece of J −i Aj . In the spin piece M −i we start with the
covariant-formalism equation M −i Aj = −δ ij A− , substitute the solution to A− ’s field
equation, and then add a gauge transformation to cancel the change of gauge induced
by the covariant-formalism transformation M −i A+ = Ai . The net result is that in
the light-cone formalism
J −i Aj = −i(x− pi − xi p− )Aj − δ ij A− + [∇j , Ai ] , (2.4.2)

with A− given by (2.1.7) and p− Aj by (2.4.1). In the abelian case, these expressions
agree with those obtained by a different method in (2.3.4). All transformations can
then be written in functional second-quantized form as

δ=− dD−2 xi dx− tr (δAi ) → [δ, Ai ] = −(δAi ) . (2.4.3)

The minus sign is as in (2.2.5) for relating first- and second-quantized operators.

As an alternative, we can consider canonical second-quantization, which has cer-
tain advantages in the light cone, and has an interesting generalization in the covariant
case (see sect. 3.4). From the light-cone lagrangian

L = −i
Φ† p+ Φ − H(Φ) , (2.4.4)

. is the “time”-derivative i∂/∂x+ , we find that the fields have equal-time
commutators similar to those in nonrelativistic field theory:
[Φ† (1), Φ(2)] = − δ(2 − 1) , (2.4.5)
2.4. Interactions 17

where the δ-function is over the transverse coordinates and x− (and may include a
Kronecker δ in indices, if Φ has components). Unlike nonrelativistic field theory, the
fields satisfy a reality condition, in coordinate space:

Φ* = ΩΦ , (2.4.6)

where Ω is the identity or some symmetric, unitary matrix (the “charge conjugation”
matrix; * here is the hermitian conjugate, or adjoint, in the operator sense, i.e., unlike

, it excludes matrix transposition). As in quantum mechanics (or the Poisson bracket
approach to classical mechanics), the generators can then be written as functions of
the dynamical variables:

V = dz1 · · · dzn V (n) (z1 , . . . , zn )Φ(z1 ) · · · Φ(zn ) , (2.4.7)
n n!

where the arguments z stand for either coordinates or momenta and the V’s are the
vertex functions, which are just functions of the coordinates (not operators). Without
loss of generality they can be chosen to be cyclically symmetric in the fields (or totally
symmetric, if group-theory indices are also permuted). (Any asymmetric piece can
be seen to contribute to a lower-point function by the use of (2.4.5,6).) In light-cone
theories the coordinate-space integrals are over all coordinates except x+ . The action
of the second-quantized operator V on fields is calculated using (2.4.5):

1 1
[V, Φ(z1 )† ] = − dz2 · · · dzn V (n) (z1 , . . . , zn )Φ(z2 ) · · · Φ(zn ) .
2p+1 n (n − 1)!
A particular case of the above equations is the free case, where the operator V is
quadratic in Φ. We will then generally write the second-quantized operator V in
terms of a first-quantized operator V with a single integration:

V = dz Φ† p+ VΦ → [V, Φ] = −VΦ . (2.4.9)

This can be checked to relate to (2.4.7) as V (2) (z1 , z2 ) = 2Ω1 p+1 V1 δ(2 − 1) (with
the symmetry of V (2) imposing corresponding conditions on the operator V). In the
interacting case, the generalization of (2.4.9) is

V = dz Φ† 2p+ (VΦ) , (2.4.10)
where N is just the number of fields in any particular term. (In the free case N = 2,
giving (2.4.9).)

For example, for Yang-Mills, we find

p− = 1
4 (F ij ) + 12 (p+ A− )2 , (2.4.11a)

J −i = ix− (p+ Aj )(pi Aj ) + ixi 1
4 (F jk ) + 12 (p+ A− )2 − Ai p+ A− . (2.4.11b)

(The other generators follow trivially from (2.4.9).) p− is minus the hamiltonian H
(as in the free case (2.1.2,4,9)), as also follows from performing the usual Legendre
transformation on the lagrangian.

In general, all the explicit xi -dependence of all the Poincaré generators can be de-
termined from the commutation relations with the momenta (translation generators)
pi . Furthermore, since only p− and J−i get contributions from interactions, we need
consider only those. Let’s first consider the “hamiltonian” p− . Since it commutes
with pi , it is translation invariant. In terms of the vertex functions, this translates
into the condition:
(p1 + · · · + pn )V (n) (p1 , . . . , pn ) = 0 , (2.4.12)

where the  indicates Fourier transformation with respect to the coordinate-space
expression, implying that most generally

V (n) (p1 , . . . , pn ) = f(p
˜ 1 , . . . , pn−1 )δ(p1 + · · · + pn ) , (2.4.13)

or in coordinate space
(n) ∂ ∂
V (x1 , . . . , xn ) = f˜ i ,...,i δ(x1 − xn ) · · · δ(xn−1 − xn )
∂x1 ∂xn−1
= f (x1 − xn , . . . , xn−1 − xn ) . (2.4.14)

In this coordinate representation one can see that when V is inserted back in (2.4.7)
we have the usual expression for a translation-invariant vertex used in field theory.
Namely, fields at the same point in coordinate space, with derivatives acting on them,
are multiplied and integrated over coordinate space. In this form it is clear that there
is no explicit coordinate dependence in the vertex. As can be seen in (2.4.14), the most
general translationally invariant vertex involves an arbitrary function of coordinate
differences, denoted as f above. For the case of bosonic coordinates, the function
f˜ may contain inverse derivatives (that is, translational invariance does not imply
locality.) For the case of anticommuting coordinates (see sect. 2.6) the situation is
simpler: There is no locality issue, since the most general function f can always be
obtained from a function f˜ polynomial in derivatives, acting on δ-functions.
2.5. Graphs 19

We now consider J−i . From the commutation relations we find:

[pi , J−j } = −ηij p− → [J−i , Φ] = ixi [p− , Φ] + [∆J−i , Φ] , (2.4.15)

where ∆J−i is translationally invariant (commutes with pi ), and can therefore be
represented without explicit xi ’s. For the Yang-Mills case, this can be seen to agree
with (2.4.2) or (2.4.11).
This light-cone analysis will be applied to interacting strings in chapt. 10.

2.5. Graphs
Feynman graphs for any interacting light-cone field theory can be derived as in
covariant field theory, but an alternative not available there is to use a nonrelativistic
style of perturbation (i.e., just expanding eiHt in H IN T ), since the field equations are
now linear in the time derivative p− = i∂/∂x+ = i∂/∂τ . (As in sect. 2.1, but unlike
sects. 2.3 and 2.4, we now use p− to refer to this partial derivative, as in covariant
formalisms, while −H refers to the corresponding light-cone Poincaré generator, the
two being equal on shell.) This formalism can be derived straightforwardly from the
usual Feynman rules (after choosing the light-cone gauge and eliminating auxiliary
fields) by simply Fourier transforming from p− to x+ = τ (but keeping all other
dp− −ip− τ 1 1 iτ (pi 2 +m2 )/2p+
e = −iΘ(p+ τ ) e . (2.5.1)
−∞ 2π 2 2
2p+ p− + pi + m + i
2|p+ |
(Θ(u) = 1 for u > 1, 0 for u < 1.) We now draw all graphs to represent the τ
coordinate, so that graphs with different τ -orderings of the vertices must be considered
as separate contributions. Then we direct all the propagators toward increasing τ , so
the change in τ between the ends of the propagator (as appears in (2.5.1)) is always
positive (i.e., the orientation of the momenta is defined to be toward increasing τ ).
We next Wick rotate τ → iτ . We also introduce external line factors which transform
H back to −p− on external lines. The resulting rules are:
(a) Assign a τ to each vertex, and order them with respect to τ .
(b) Assign (p− , p+ , pi ) to each external line, but only (p+ , pi ) to each internal line, all
directed toward increasing τ . Enforce conservation of (p+ , pi ) at each vertex, and
total conservation of p− .
(c) Give each internal line a propagator
1 −τ (pi 2 +m2 )/2p+
Θ(p+ ) e

for the (p+ , pi ) of that line and the positive difference τ in the proper time between
the ends.

(d) Give each external line a factor
eτ p−

for the p− of that line and the τ of the vertex to which it connects.

(e) Read off the vertices from the action as usual.

(f) Integrate  ∞


for each τ difference between consecutive (though not necessarily connected) ver-
tices. (Performing just this integration gives the usual old-fashioned perturbation
theory in terms of energy denominators [2.1], except that our external-line factors
differ off shell in order to reproduce the usual Feynman rules.)

(g) Integrate
dp+ dD−2pi
−∞ (2π)D−1
for each loop.

The use of such methods for strings will be discussed in chapt. 10.

2.6. Covariantized light cone

There is a covariant formalism for any field theory that has the interesting prop-
erty that it can be obtained directly and easily from the light-cone formalism, without
any additional gauge-fixing procedure [2.7]. Although this covariant gauge is not as
general or convenient as the usual covariant gauges (in particular, it sometimes has
additional off-shell infrared divergences), it bears strong relationship to both the light-
cone and BRST formalisms, and can be used as a conceptual bridge. The basic idea
of the formalism is: Consider a covariant theory in D dimensions. This is equivalent
to a covariant theory in (D + 2) − 2 dimensions, where the notation indicates the ad-
dition of 2 extra commuting coordinates (1 space, 1 time) and 2 (real) anticommuting
coordinates, with a similar extension of Lorentz indices [2.8]. (A similar use of OSp
groups in gauge-fixed theories, but applied to only the Lorentz indices and not the co-
ordinates, appears in [2.9].) This extends the Poincaré group ISO(D−1,1) to a graded
analog IOSp(D,2|2). In practice, this means we just take the light-cone transverse in-
dices to be graded, watching out for signs introduced by the corresponding change in
2.6. Covariantized light cone 21

statistics, and replace the Euclidean SO(D-2) metric with the corresponding graded
OSp(D-1,1|2) metric:

i = (a, α) , δ ij → η ij = (η ab , C αβ ) , (2.6.1)

where η ab is the usual Lorentz metric and

C αβ = C βα = σ 2 (2.6.2)

is the Sp(2) metric, which satisfies the useful identity

C αβ C γδ = δ [α γ δ β] δ → A[α B β] = C αβ C γδ Aγ B δ . (2.6.3)

The OSp metric is used to raise and lower graded indices as:

xi = η ij xj , xi = xj η ji ; η ik η jk = δ j i . (2.6.4)

The sign conventions are that adjacent indices are contracted with the contravariant
(up) index first. The equivalence follows from the fact that, for momentum-space
Feynman graphs, the trees will be the same if we constrain the 2 − 2 extra “ghost”
momenta to vanish on external lines (since they’ll then vanish on internal lines by
momentum conservation); and the loops are then the same because, when the mo-
mentum integrands are written as gaussians, the determinant factors coming from the
2 extra anticommuting dimensions exactly cancel those from the 2 extra commuting
ones. For example, using the proper-time form (“Schwinger parametrization”) of the
propagators (cf. (2.5.1)),
1 2 +m2 )
= dτ e−τ (p , (2.6.5)
p + m2
2 0

all momentum integrations take the form

1  D+2 2 −f (2p+ p− +pa pa +pα pα +m2 )

= dD p e−f (p pa +m )
d p d pα e
π 2
= e−f m , (2.6.6)

where f is a function of the proper-time parameters.
The covariant theory is thus obtained from the light-cone one by the substitution

(p− , p+ ; pi ) → (p− , p+ ; pa , pα ) , (2.6.7a)

p− = pα = 0 (2.6.7b)

on physical states. It’s not necessary to set p+ = 0, since it only appears in the combi-
nation p− p+ in OSp(D,2|2)-invariant products. Thus, p+ can be chosen arbitrarily on
external lines (but should be nonvanishing due to the appearance of factors of 1/p+ ).
We now interpret x± and xα as the unphysical coordinates. Vector indices on fields
are treated similarly: Having been reduced to transverse ones by the light-cone for-
malism, they now become covariant vector indices with 2 additional anticommuting
values ((2.6.1)). For example, in Yang-Mills the vector field becomes the usual vector
field plus two anticommuting scalars Aα , corresponding to Faddeev-Popov ghosts.

The graphical rules become:

(a) Assign a τ to each vertex, and order them with respect to τ .

(b) Assign (p+ , pa ) to each external line, but (p+ , pa , pα ) to each internal line, all
directed toward increasing τ . Enforce conservation of (p+ , pa , pα ) at each vertex
(with pα = 0 on external lines).

(c) Give each internal line a propagator

1 −τ (pa 2 +pαpα +m2)/2p+
Θ(p+) e

for the (p+ , pa , pα ) of that line and the positive difference τ in the proper time between
the ends.

(d) Give each external line a factor
1 .

(e) Read off the vertices from the action as usual.

(f) Integrate  ∞


for each τ difference between consecutive (though not necessarily connected) ver-

(g) Integrate 
d 2 pα
Exercises 23

for each loop (remembering that for any anticommuting variable θ, dθ 1 = 0,
dθ θ = 1, θ = 0).

(h) Integrate  ∞
2 dp+

for each loop.

(i) Integrate

dD p
for each loop.

For theories with only scalars, integrating just (f-h) gives the usual Feynman
graphs (although it may be necessary to add several graphs due to the τ -ordering of
non-adjacent vertices). Besides the correspondence of the τ parameters to the usual
Schwinger parameters, after integrating out just the anticommuting parameters the
p+ parameters resemble Feynman parameters.

These methods can also be applied to strings (chapt. 10).


(1) Find the light-cone formulation of QED. Compare with the Coulomb gauge for-

(2) Derive the commutation relations of the conformal group from (2.2.2). Check
that (2.2.4) satisfies them. Evaluate the commutators implicit in (2.2.7) for each

(3) Find the Lorentz transformation M ab of a vector (consistent with the conventions
of (2.2.2)). (Hint: Look at the transformations of x and p.) Find the explicit
form of (2.2.8) for that case. Solve these equations of motion. To what simpler
representation is this equivalent? Study this equivalence with the light-cone anal-
ysis given below (2.2.8). Generalize the analysis to totally antisymmetric tensors
of arbitrary rank.

(4) Repeat problem (3) for the massive case. Looking at the separate SO(D-1,1)
representations contained in the SO(D,1) representations, show that first-order
formalisms in terms of the usual fields have been obtained, and find the corre-
sponding second-order formulations.

(5) Check that the explicit forms of the Poincaré generators given in (2.3.5) satisfy
the correct algebra (see problem (2)). Find the explicit transformations acting
on the vector representation of the spin group SO(D-1). Compare with (2.4.1-2).

(6) Derive (2.4.11). Compare that p− with the light-cone hamiltonian which follows
from (2.1.5).

(7) Calculate the 4-point amplitude in φ3 theory with light-cone graphs, and com-
pare with the usual covariant Feynman graph calculation. Calculate the 1-loop
propagator correction in the same theory using the covariantized light-cone rules,
and again compare with ordinary Feynman graphs, paying special attention to
Feynman parameters.
3.1. Gauge invariance and constraints 25


3.1. Gauge invariance and constraints

In the previous chapter we saw that a gauge theory can be described either in a
manifestly covariant way by using gauge degrees of freedom, or in a manifestly unitary
way (with only physical degrees of freedom) with Poincaré transformations which are
nonlinear (in both coordinates and fields). In the gauge-covariant formalism there is a
D-dimensional manifest Lorentz covariance, and in the light-cone formalism a D − 2-
dimensional one, and in each case a corresponding number of degrees of freedom.
There is also an intermediate formalism, more familiar from nonrelativistic theory:
The hamiltonian formalism has a D − 1-dimensional manifest Lorentz covariance (ro-
tations). As in the light-cone formalism, the notational separation of coordinates
into time and space suggests a particular type of gauge condition: temporal (time-
like) gauges, where time-components of gauge fields are set to vanish. In chapt. 5,
this formalism will be seen to have a particular advantage for first-quantization of
relativistic theories: In the classical mechanics of relativistic theories, the coordinates
are treated as functions of a “proper time” so that the usual time coordinate can be
treated on an equal footing with the space coordinates. Thus, canonical quantization
with respect to this unobservable (proper) “time” coordinate doesn’t destroy manifest
Poincaré covariance, so use of a hamiltonian formalism can be advantageous, partic-
ularly in deriving BRST transformations, and the corresponding second-quantized
theory, where the proper-time doesn’t appear anyway.
We’ll first consider Yang-Mills, and then generalize to arbitrary gauge theories.
In order to study the temporal gauge, instead of the decomposition (2.1.1) we simply
separate into time and spatial components

a = (0, i) , A · B = −A0 B 0 + Ai B i . (3.1.1)

The lagrangian (2.1.5) is then

L = 14 F ij 2 − 12 (p0 Ai − [∇i , A0 ])2 . (3.1.2)

The gauge condition
A0 = 0 (3.1.3)
transforms under a gauge transformation with a time derivative: Under an infinites-
imal transformation about A0 = 0,

δA0 ≈ ∂ 0 λ , (3.1.4)

so the Faddeev-Popov ghosts are propagating. Furthermore, the gauge transformation
(3.1.4) does not allow the gauge choice (3.1.3) everywhere: For example, if we choose
periodic boundary conditions in time (to simplify the argument), then
δ dx0 A0 ≈ 0 . (3.1.5)

A0 can then be fixed by an appropriate initial condition, e.g., A0 |x0 =0 = 0, but then
the corresponding field equation is lost. Therefore, we must impose
0= = −[∇i , F 0i ] = −[∇i , p0 Ai ] at x0 = 0 (3.1.6)
as an initial condition. Another way to understand this is to note that gauge fixing
eliminates only degrees of freedom which don’t occur in the lagrangian, and thus
can eliminate only redundant equations of motion: Since [∇i , F 0i ] = 0 followed from
the gauge-invariant action, the fact that it doesn’t follow after setting A0 = 0 means
some piece of A0 can’t truly be gauged away, and so we must compensate by imposing
the equation of motion for that piece. Due to the original gauge invariance, (3.1.6)
then holds for all time from the remaining field equations: In the gauge (3.1.3), the
lagrangian (3.1.2) becomes

L = 12 Ai 2 Ai − 12 (pi Ai )2 + [Ai , Aj ]pi Aj + 14 [Ai , Aj ]2 , (3.1.7)

and the covariant divergence of the implied field equations yields the time derivative
of (3.1.6). (This follows from the identity [∇b , [∇a , F ab ]] = 0 upon applying the
field equations [∇a , F ia ] = 0. In unitary gauges, the corresponding constraint can be
derived without time derivatives, and hence is implied by the remaining field equations
under suitable boundary conditions.) Equivalently, if we notice that (3.1.4) does not
fix the gauge completely, but leaves time-independent gauge transformations, we need
to impose a constraint on the initial states to make them gauge invariant. But the
generator of the residual gauge transformations on the remaining fields Ai is
G(xi ) = ∇i , i , (3.1.8)
3.1. Gauge invariance and constraints 27

which is the same as the constraint (3.1.6) under canonical quantization of (3.1.7).
Thus, the same operator (1) gives the constraint which must be imposed in addition to
the field equations because too much of A0 was dropped, and (2) (its transpose) gives
the gauge transformations remaining because they left the gauge-fixing function A0
invariant. The fact that these are identical is not surprising, since in Faddeev-Popov
quantization the latter corresponds to the Faddeev-Popov ghost while the former
corresponds to the antighost.

These properties appear very naturally in a hamiltonian formulation: We start

again with the gauge-invariant lagrangian (3.1.2). Since A0 has no time-derivative
terms, we Legendre transform with respect to just Ai . The result is

= 2

dD x tr LH ,
LH = Ai Πi − H , H = H0 + A0 iG ,

H0 = 12 Πi 2 − 14 F ij 2 , G = [∇i , Πi ] , (3.1.9)

. = ∂ 0 . As in ordinary nonrelativistic classical mechanics, eliminating the
momentum Πi from the hamiltonian form of the action (first order in time deriva-
tives) by its equation of motion gives back the lagrangian form (second order in time
derivatives). Note that A0 appears linearly, as a Lagrange multiplier.

The gauge-invariant hamiltonian formalism of (3.1.9) can be generalized [3.1]:
Consider a lagrangian of the form
LH = z M eM A (z)π A − H , H = H0 (z, π) + λi iG i (z, π) , (3.1.10)

where z, π, and λ are the variables, representing “coordinates,” covariant “momenta,”
and Lagrange multipliers, respectively. They depend on the time, and also have
indices (which may include continuous indices, such as spatial coordinates). e, which
is a function of z, has been introduced to allow for cases with a symmetry (such
as supersymmetry) under which dz M eM A (but not dz itself) is covariant, so that π
will be covariant, and thus a more convenient variable in terms of which to express
the constraints G. When H0 commutes with G (quantum mechanically, or in terms
of Poisson brackets for a classical treatment), this action has a gauge invariance
generated by G, for which λ is the gauge field:

δ(z, π) = [ζ iG i , (z, π)] ,


− λi G i = 0

(δλi )G i = ζ i G i + [λj G j , ζ iG i ] , (3.1.11)

where the gauge transformation of λ has been determined by the invariance of the
“total” time-derivative d/dt = ∂/∂t + iH. (More generally, if [ζ iG i , H0 ] = f i iG i , then
δλi has an extra term −f i .) Using the chain rule ((d/dt) on f (t, q k (t)) equals ∂/∂t +
q k (∂/∂q k )) to evaluate the time derivative of G, we find the lagrangian transforms as
a total derivative
d M
δLH = (δz )eM A π A − ζ i iG i , (3.1.12)
which is the usual transformation law for an action with local symmetry generated
by the current G. When H0 vanishes (as in relativistic mechanics), the special case
ζ i = ζλi of the transformations of (3.1.11) are τ reparametrizations, generated by the
hamiltonian λi G i . In general, after canonical quantization, the wave function satisfies
the Schrödinger equation ∂/∂t + iH0 = 0, as well as the constraints G = 0 (and thus
∂/∂t + iH = 0 in any gauge choice for λ). Since [H0 , G] = 0, G = 0 at t = 0 implies
G = 0 for all t.

In some cases (such as Yang-Mills), the Lorentz covariant form of the action can
be obtained by eliminating all the π’s. A covariant first-order form can generally be
obtained by introducing additional auxiliary degrees of freedom which enlarge π to
make it Lorentz covariant. For example, for Yang-Mills we can rewrite (3.1.9) as

LH = 12 G0i 2 − G0 i F 0i + 14 F ij 2

→ L1 = − 14 Gab 2 + Gab F ab , (3.1.13)

where G0i = iΠi , and the independent (auxiliary) fields Gab also include Gij , which
have been introduced to put 14 F ij 2 into first-order form and thus make the lagrangian
manifestly Lorentz covariant. Eliminating Gij by their field equations gives back the
hamiltonian form.

Many examples will be given in chapts. 5-6 for relativistic first-quantization,
where H0 vanishes, and thus the Schrödinger equation implies the wave function is
proper-time-independent (i.e., we require H0 = 0 because the proper time is not
physically observable). Here we give an interesting example in D=2 which will also
be useful for strings. Consider a single field A with canonical momentum P and
iG = 14 (P + A )2 , H0 = 14 (P − A )2 , (3.1.14)

where is the derivative with respect to the 1 space coordinate (which acts as the
index M or i from above). From the algebra of P ± A , it’s easy to check, at least
at the Poisson bracket level, that the G algebra closes and H0 is invariant. (This
3.2. IGL(1) 29

algebra, with particular boundary conditions, will be important in string theory: See
chapt. 8. Note that P + A does not form an algebra, so its square must be used.)
The transformation laws (3.1.11) are found to be

δA = ζ 12 (P + A ) ,
. ←→
δλ = ζ − λ ∂ 1 ζ . (3.1.15)

In the gauge λ = 1 the action becomes the usual hamiltonian one for a massless
scalar, but the constraint implies P + A = 0, which means that modes propagate
only to the right and not the left. The lagrangian form again results from eliminating
P , and after the redefinitions
1−λ √ 1
λ̂ = 2 , ζ̂ = 2 ζ , (3.1.16)
1+λ 1+λ
we find [3.2]
L = −(∂ + A)(∂ − A) + 12 λ̂(∂ − A)2 ;
δA = ζ̂∂ − A , δ λ̂ = 2∂ + ζ̂ + ζ̂ ∂ − λ̂ ; (3.1.17)

where ∂ ± are defined as in sect. 2.1.

The gauge fixing (including Faddeev-Popov ghosts) and initial condition can be
described in a very concise way by the BRST method. The basic idea is to construct
a symmetry relating the Faddeev-Popov ghosts to the unphysical modes of the gauge
field. For example, in Yang-Mills only D − 2 Lorentz components of the gauge field
are physical, so the Lorentz-gauge D-component gauge field requires 2 Faddeev-Popov
ghosts while the temporal-gauge D − 1-component field requires only 1. The BRST
symmetry rotates the additional gauge-field components into the FP ghosts, and vice
versa. Since the FP ghosts are anticommuting, the generator of this symmetry must
be, also.

3.2. IGL(1)

We will find that the methods of Becchi, Rouet, Stora, and Tyutin [3.3] are the
most useful way not only to perform quantization in Lorentz-covariant and general
nonunitary gauges, but also to derive gauge-invariant theories. BRST quantization is
a more general way of quantizing gauge theories than either canonical or path-integral
(Faddeev-Popov), because it (1) allows more general gauges, (2) gives the Slavnov-
Taylor identities (conditions for unitarity) directly (they’re just the Ward identities
for BRST invariance), and (3) can separate the gauge-invariant part of a gauge-fixed

action. It is defined by the conditions: (1) BRST transformations form a global group
with a single (abelian) anticommuting generator Q. The group property then implies
Q2 = 0 for closure. (2) Q acts on physical fields as a gauge transformation with the
gauge parameter replaced by the (real) ghost. (3) Q on the (real) antighost gives a
BRST auxiliary field (necessary for closure of the algebra off shell). Nilpotence of Q
then implies that the auxiliary field is BRST invariant. Physical states are defined
to be those which are BRST invariant (modulo null states, which can be expressed
as Q on something) and have vanishing ghost number (the number of ghosts minus

There are two types of BRST formalisms: (1) first-quantized-style BRST, origi-
nally found in string theory [3.4] but also applicable to ordinary field theory, which
contains all the field equations as well as the gauge transformations; and (2) second-
quantized-style BRST, the original form of BRST, which contains only the gauge
transformations, corresponding in a hamiltonian formalism to those field equations
(constraints) found from varying the time components of the gauge fields. However,
we’ll find (in sect. 4.4) that, after restriction to a certain subset of the fields, BRST1 is
equivalent to BRST2. (It’s the BRST variation of the additional fields of BRST1 that
leads to the field equations for the physical fields.) The BRST2 transformations were
originally found from Yang-Mills theory. We will first derive the YM BRST2 transfor-
mations, and by a simple generalization find BRST operators for arbitrary theories,
applicable to BRST1 or BRST2 and to lagrangian or hamiltonian formalisms.

In the general case, there are two forms for the BRST operators, correspond-
ing to different classes of gauges. The gauges commonly used in field theory fall
into three classes: (1) unitary (Coulomb, Arnowitt-Fickler/axial, light-cone) gauges,
where the ghosts are nonpropagating, and the constraints are solved explicitly (since
they contain no time derivatives); (2) temporal/timelike gauges, where the ghosts have
equations of motion first-order in time derivatives (making them canonically conju-
gate to the antighosts); and (3) Lorentz (Landau, Fermi-Feynman) gauges, where
the ghost equations are second-order (so ghosts are independent of antighosts), and
the Nakanishi-Lautrup auxiliary fields [3.5] (Lagrange multipliers for the gauge con-
ditions) are canonically conjugate to the auxiliary time-components of the gauge
fields. Unitary gauges have only physical polarizations; temporal gauges have an
additional pair of unphysical polarizations of opposite statistics for each gauge gener-
ator; Lorentz gauges have two pairs. In unitary gauges the BRST operator vanishes
identically; in temporal gauges it is constructed from group generators, or constraints,
3.2. IGL(1) 31

multiplied by the corresponding ghosts, plus terms for nilpotence; in Lorentz gauges
it has an extra “abelian” term consisting of the products of the second set of unphys-
ical fields. Temporal-gauge BRST is defined in terms of a ghost number operator in
addition to the BRST operator, which itself has ghost number 1. We therefore refer
to this formalism by the corresponding symmetry group with two generators, IGL(1).
Lorentz-gauge BRST has also an antiBRST operator [3.6], and this and BRST trans-
form as an “isospin” doublet, giving the larger group ISp(2), which can be extended
further to OSp(1,1|2) [2.3,3.7]. Although the BRST2 OSp operators are generally
of little value (only the IGL is required for quantization), the BRST1 OSp gives a
powerful method for obtaining free gauge-invariant formalisms for arbitrary (particle
or string) field theories. In particular, for arbitrary representations of the Poincaré
group a certain OSp(1,1|2) can be extended to IOSp(D,2|2) [2.3], which is derived
from (but does not directly correspond to quantization in) the light-cone gauge.

One simple way to formulate anticommuting symmetries (such as supersymme-
try) is through the use of anticommuting coordinates [3.8]. We therefore extend
spacetime to include one extra, anticommuting coordinate, corresponding to the one
anticommuting symmetry:
a → (a, α) (3.2.1)

for all vector indices, including those on coordinates, with Fermi statistics for all
quantities with an odd number of anticommuting indices. (α takes only one value.)
Covariant derivatives and gauge transformations are then defined by the correspond-
ing generalization of (2.1.5b), and field strengths with graded commutators (commu-
tators or anticommutators, according to the statistics). However, unlike supersym-
metry, the extra coordinate does not represent extra physical degrees of freedom, and
so we constrain all field strengths with anticommuting indices to vanish [3.9]: For
F αa = F αβ = 0 , (3.2.2a)

so that gauge-invariant quantities can be constructed only from the usual F ab . When
Yang-Mills is coupled to matter fields φ, we similarly have the constraints

∇ α φ = ∇α ∇a φ = 0 , (3.2.2b)

and these in fact imply (3.2.2a) (consider {∇α , ∇β } and [∇α , ∇a ] acting on φ). These
constraints can be solved easily:

F αa = 0 → pα Aa = [∇a , Aα ] ,

F αβ = 0 → pα Aβ = − 12 {Aα , Aβ } = −Aα Aβ ;

∇α φ = 0 → pα φ = −Aα φ . (3.2.3)

(In the second line we have used the fact that α takes only one value.) Defining “ | ”
to mean |xα =0 , we now interpret Aa | as the usual gauge field, iAα | as the FP ghost,
and the BRST operator Q as Q(ψ|) = (pα ψ)|. (Similarly, φ| is the usual matter
field.) Then ∂ α ∂ β = 0 (since α takes only one value and ∂ α is anticommuting) implies
Q2 = 0 . (3.2.4)

In a hamiltonian approach [3.10] these transformations are sufficient to perform quan-
tization in a temporal gauge, but for the lagrangian approach or Lorentz gauges we
also need the FP antighost and Nakanishi-Lautrup auxiliary field, which we define in
terms of an unconstrained scalar field A:  is the antighost, and

B = (pα iA)| (3.2.5)

is the auxiliary field.
The BRST transformations (3.2.3) can be represented in operator form as
∂ ∂
Q = C i G i + 12 C j C i f ij k k
− iB i  , (3.2.6a)
∂C ∂C i
where i is a combined space(time)/internal-symmetry index, C is the FP ghost, C is
the FP antighost, B is the NL auxiliary field, and the action on the physical fields is
given by the constraint/gauge-transformation G satisfying the algebra

[G i , G j } = f ij k G k , (3.2.6b)

where we have generalized to graded algebras with graded commutator [ , } (com-
mutator or anticommutator, as appropriate). In this case,
G = ∇, ·i , (3.2.7)

where the structure constants in (3.2.6b) are the usual group structure constants
times δ-functions in the coordinates. Q of (3.2.6a) is antihermitian when C, C,
B are hermitian and G is antihermitian, and is nilpotent (3.2.4) as a consequence of
(3.2.6b). Since C and B appear only in the last term in (3.2.6a), these properties
also hold if that term is dropped. (In the notation of (3.2.1-5), the fields A and A are
3.2. IGL(1) 33

When [G i , f jk l } = 0, (3.2.6a) still gives Q2 = 0. However, when the gauge
invariance has a gauge invariance of its own, i.e., Λi G i = 0 for some nontrivial Λ
depending on the physical variables implicit in G, then, although (3.2.6a) is still
nilpotent, it requires extra terms in order to allow gauge fixing this invariance of the
ghosts. In some cases (see sect. 5.4) this requires an infinite number of new terms (and
ghosts). In general, the procedure of adding in the additional ghosts and invariances
can be tedious, but in sect. 3.4 we’ll find a method which automatically gives them
all at once.
The gauge-fixed action is required to be BRST-invariant. The gauge-invariant
part already is, since Q on physical fields is a special case of a gauge transforma-
tion. The gauge-invariant lagrangian is quantized by adding terms which are Q on
something (corresponding to integration over xα ), and thus BRST-invariant (since
Q2 = 0): For example, rewriting (3.2.3,5) in the present notation,

QAa = −i[∇a , C] ,

QC = iC 2 ,

QC̃ = −iB ,

QB = 0 , (3.2.8)

we can choose
LGF = iQ C [f (A) + g(B)] = B [f (A) + g(B)] − C [∇a , C] , (3.2.9)
which gives the usual FP term for gauge condition f (A) = 0 with gauge-averaging
function Bg(B). However, gauges more general than FP can be obtained by putting
more complicated ghost-dependence into the function on which Q acts, giving terms
more than quadratic in ghosts. In the temporal gauge

f (A) = A0 (3.2.10)

and g contains no time derivatives in (3.2.9), so upon quantization B is eliminated
(it’s nonpropagating) and C is canonically conjugate to C. Thus, in the hamiltonian
formalism (3.2.6a) gives the correct BRST transformations without the last term,
where the fields are now functions of just space and not time, the sum in (3.2.7) runs
over just the spatial values of the spacetime index as in (3.1.8), and the derivatives
correspond to functional derivatives which give δ functions in just spatial coordinates.
On the other hand, in Lorentz gauges the ghost and antighost are independent even

after quantization, and the last term in Q is needed in both lagrangian and hamil-
tonian formalisms; but the product in (3.2.7) and the arguments of the fields and
δ functions are as in the temporal gauge. Therefore, in the lagrangian approach Q
is gauge independent, while in the hamiltonian approach the only gauge dependence
is the set of unphysical fields, and thus the last term in Q. Specifically, for Lorentz
gauges we choose
f (A) = ∂ · A , g(B) = 12 ζB →

LGF = ζ 12 B 2 + B∂ · A − C∂ · [∇, C]

= − 12 (∂ · A)2 + ζ 12 B 2 − C∂
 · [∇, C] ,

 = B + 1∂ · A ,
B (3.2.11)
using (3.2.9).

The main result is that (3.2.6a) gives a general BRST operator for arbitrary alge-
bras (3.2.6b), for hamiltonian or lagrangian formalisms, for arbitrary gauges (includ-
ing temporal and Lorentz), where the last term contains arbitrary numbers (perhaps
 B) fields. Since G = 0 is the field equation (3.1.6), physical states
0) of sets of (C,
must satisfy Qψ = 0. Actually, G = 0 is satisfied only as a Gupta-Bleuler condi-
tion, but still Qψ = 0 because in the C i G i term in (3.2.6a) positive-energy parts
of C i multiply negative-energy parts of G i , and vice versa. Thus, for any value of
an appropriate index i, either C i |ψ
= ψ| G i = 0 or G i |ψ
= ψ| C i = 0, modulo
contributions from the C 2 ∂/∂C term. However, since G is also the generator of gauge
transformations (3.1.8), any state of the form ψ + Qλ is equivalent to ψ. The physical
states are therefore said to belong to the “cohomology” of Q: those satisfying Qψ = 0
modulo gauge transformations δψ = Qλ. (“Physical” has a more restrictive meaning
in BRST1 than BRST2: In BRST2 the physical states are just the gauge-invariant
ones, while in BRST1 they must also be on shell.) In addition, physical states must
have a specified value of the ghost number, defined by the ghost number operator
∂ ∂
J3 = Ci i
− C i  , (3.2.12a)
∂C ∂C i
[J 3 , Q] = Q , (3.2.12b)

and the latter term in (3.2.12a) is dropped if the last term in (3.2.6a) is. The two
operators Q and J 3 form the algebra IGL(1), which can be interpreted as a translation
3.3. OSp(1,1|2) 35

and scale transformation, respectively, with respect to the coordinate xα (i.e., the
conformal group in 1 anticommuting dimension).

From the gauge generators G i , which act on only the physical variables, we can
define IGL(1)-invariant generalizations which transform also C, as the adjoint repre-
∂ ∂
Gi = Q, = G i + C j f jik . (3.2.13)
∂C i ∂C k
 are gauge-fixed versions of the gauge generators G.
The G’s

Types of gauges for first-quantized theories will be discussed in chapt. 5 for par-
ticles and chapt. 6 and sect. 8.3 for strings. Gauge fixing for general field theories
using BRST will be described in sect. 4.4, and for closed string field theory in sect.
11.1. IGL(1) algebras will be used for deriving general gauge-invariant free actions in
sect. 4.2. The algebra will be derived from first-quantization for the particle in sect.
5.2 and for the string in sect. 8.1. However, in the next section we’ll find that IGL(1)
can always be derived as a subgroup of OSp(1,1|2), which can be derived in a more
general way than by first-quantization.

3.3. OSp(1,1|2)

Although the IGL(1) algebra is sufficient for quantization in arbitrary gauges, in
the following section we will find the larger OSp(1,1|2) algebra useful for the BRST1
formalism, so we give a derivation here for BRST2 and again generalize to arbitrary
BRST. The basic idea is to introduce a second BRST, “antiBRST,” corresponding to
the antighost. We therefore repeat the procedure of (3.2.1-7) with 2 anticommuting
coordinates [3.11] by simply letting the index α run over 2 values (cf. sect. 2.6). The
solution to (3.2.2) is now

F αa = 0 → pα Aa = [∇a , Aα ] ,

F αβ = 0 → pα Aβ = − 12 {Aα , Aβ } − iC αβ B ;

∇α φ = 0 → pα φ = −Aα φ ; (3.3.1a)

where Aα now includes both ghost and antighost. The appearance of the NL field is
due to the ambiguity in the constraint F αβ = p(α Aβ) +· · ·. The remaining (anti)BRST
transformation then follows from further differentiation:

{pα , pβ }Aγ = 0 → pα B = − 12 [Aα , B] + i 12
Aβ , {Aα , Aβ } . (3.3.1b)

The generalization of (3.2.6a) is then [3.12], defining Qα (ψ|) = (∂ α ψ)| (and renaming
C α = Aα ),
∂ ∂ ∂
Qα = C iα G i + 12 C jα C iβ f ij k kβ
− B i i + 12 C jα B i f ij k
∂C ∂C α ∂B k

1 ∂
− 12 C kβ C jαC i β f ij l f lk m , (3.3.2)
∂B m
and of (3.2.12a) is

J αβ = C i (α
, (3.3.3)
∂C iβ)
where ( ) means index symmetrization. These operators form an ISp(2) algebra
consisting of the translations Qα and rotations J αβ on the coordinates xα :

{Qα , Qβ } = 0 ,

[J αβ , Qγ ] = −C γ(α Qβ) , [J αβ , J γδ ] = −C (γ(α J β)δ) . (3.3.4)

In order to relate to the IGL(1) formalism, we write
α  α  αβ
J+ −iJ 3
Q = (Q, Q) , C = (C, C) , J = , (3.3.5)
−iJ 3 J−
and make the unitary transformation

ln U = − 12 C j C i f ij k i . (3.3.6)
∂B k
Then UQU −1 is Q of (3.2.6a) and UJ 3 U −1 = J 3 is J 3 of (3.2.12a). However, whereas
there is an arbitrariness in the IGL(1) algebra in redefining J 3 by a constant, there
is no such ambiguity in the OSp(1,1|2) algebra (since it is “simple”).

Unlike the IGL case, the NL fields now are an essential part of the algebra.
Consequently, the algebra can be enlarged to OSp(1,1|2) [3.7]:

J −α = Qα , J +α = 2C i α ,
∂B i
∂ ∂ iα ∂
J αβ = C i (α , J −+ = 2B i + C , (3.3.7)
∂C iβ) ∂B i ∂C iα
with Qα as in (3.3.2), satisfy

[J αβ , J γδ ] = −C (γ(α J β)δ) ,

[J αβ , J ±γ ] = −C γ(α J ±β) ,

{J −α , J +β } = −C αβ J −+ − J αβ ,
3.3. OSp(1,1|2) 37

[J −+ , J ±α ] = ∓J ±α ,

rest = 0 . (3.3.8)

This group is the conformal group for xα , with the ISp(2) subgroup being the corre-
sponding Poincaré (or Euclidean) subgroup:

J −α = −∂ α , J +α = 2xβ 2 ∂ α + xβ M βα + xα d

J αβ = x(α ∂ β) + M αβ , J −+ = −xα ∂ α + d . (3.3.9)

(We define the square of an Sp(2) spinor as (xα )2 ≡ 12 xα xα .) J −α are the transla-
tions, J αβ the Lorentz transformations (rotations), J −+ the dilatations, and J +α the
conformal boosts. As a result of constraints analogous to (3.3.1a), the translations
are realized nonlinearly in (3.3.7) instead of the boosts. This should be compared
with the usual conformal group (2.2.4). The action of the generators (3.3.9) have
been chosen to have the opposite sign of those of (3.3.8), since it is a coordinate rep-
resentation instead of a field representation (see sect. 2.2). In later sections we will
actually be applying (3.3.7) to coordinates, and hence (3.3.9) should be considered a
“zeroth-quantized” formalism.

From the gauge generators G i , we can define OSp(1,1|2)-invariant generalizations
which transform also C and B, as adjoint representations:
∂ ∂ ∂
Gi = 1
2 J − α , J −α , = G i + C jα f jik + B j f jik . (3.3.10)
∂B i ∂C kα ∂B k

 are the OSp(1,1|2) generalization of the operators (3.2.13).
The G’s

The OSp(1,1|2) algebra (3.3.7) can be extended to an inhomogeneous algebra
IOSp(1,1|2) when one of the generators, which we denote by G 0 , is distinguished
[3.13]. We then define

p+ = −2i ,
∂B 0
1 ∂ ∂
pα = i + 1 C i α f i0 j ,
p+ ∂C 0α 2 ∂B j

p− =− iG 0 + pα 2 . (3.3.11)

(The i indices still include the value 0.) G0 is then the IOSp(1,1|2) invariant i 12 (2p+ p− +
pα pα ). This algebra is useful for constructing gauge field theory for closed strings.

OSp(1,1|2) will play a central role in the following chapters: In chapt. 4 it will
be used to derive free gauge-invariant actions. A more general form will be derived
in the following sections, but the methods of this section will also be used in sect. 8.3
to describe Lorentz-gauge quantization of the string.

3.4. From the light cone
In this section we will derive a general OSp(1,1|2) algebra from the light-cone
Poincaré algebra of sect. 2.3, using concepts developed in sect. 2.6. We’ll use this
general OSp(1,1|2) to derive a general IGL(1), and show how IGL(1) can be extended
to include interactions.
The IGL(1) and OSp(1,1|2) algebras of the previous section can be constructed
from an arbitrary algebra G, whether first-quantized or second-quantized, and la-
grangian or hamiltonian. That already gives 8 different types of BRST formalisms.
Furthermore, arbitrary gauges, more general than those obtained by the FP method,
and graded algebras (where some of the G’s are anticommuting, as in supersym-
metry) can be treated. However, there is a ninth BRST formalism, similar to the
BRST1 OSp(1,1|2) hamiltonian formalism, which starts from an IOSp(D,2|2) algebra
[2.3] which contains the OSp(1,1|2) as a subgroup. This approach is unique in that,
rather than starting from the gauge covariant formalism to derive the BRST algebra,
it starts from just the usual Poincaré algebra and derives both the gauge covariant
formalism and BRST algebra. In this section, instead of deriving BRST1 from first-
quantization, we will describe this special form of BRST1, and give the OSp(1,1|2)
subalgebra of which special cases will be found in the following chapters.
The basic idea of the IOSp formalism is to start from the light-cone formalism
of the theory with its nonlinear realization of the usual Poincaré group ISO(D-1,1)
(with manifest subgroup ISO(D-2)), extend this group to IOSp(D,2|2) (with manifest
IOSp(D-1,1|2)) by adding 2 commuting and 2 anticommuting coordinates, and take
the ISO(D-1,1)⊗OSp(1,1|2) subgroup, where this ISO(D-1,1) is now manifest and the
nonlinear OSp(1,1|2) is interpreted as BRST. Since the BRST operators of BRST1
contain all the field equations, the gauge-invariant action can be derived. Thus, not
only can the light-cone formalism be derived from the gauge-invariant formalism,
but the converse is also true. Furthermore, for general field theories the light-cone
formalism (at least for the free theory) is easier to derive (although more awkward
to use), and the IOSp method therefore provides a convenient method to derive the
gauge covariant formalism.
3.4. From the light cone 39

We now perform dimensional continuation as in sect. 2.6, but set x+ = 0 as in sect.
2.3. Our fields are now functions of (xa , xα , x− ), and have indices corresponding to
representations of the spin subgroup OSp(D−1,1|2) in the massless case or OSp(D,1|2)
in the massive. Of the full group IOSp(D,2|2) (obtained from extending (2.3.5)) we
are now only interested in the subgroup ISO(D−1,1)⊗OSp(1,1|2). The former factor
is the usual Poincaré group, acting only in the physical spacetime directions:

pa = i∂ a , J ab = −ix[a pb] + M ab . (3.4.1)

The latter factor is identified as the BRST group, acting in only the unphysical

J αβ = −ix(α pβ) + M αβ , J −+ = −ix− p+ + k , J +α = ixα p+ ,
J −α = −ix− pα + −ixα 12 (pb pb + M 2 + pβ pβ ) + M α β pβ + kpα + Qα ,
{Qα , Qβ } = −M αβ (pa pa + M 2 ) ; (3.4.2a)

Qα = M α b pb + M αm M . (3.4.2b)

We’ll generally set k = 0.

In order to relate to the BRST1 IGL formalism obtained from ordinary first-
quantization (and discussed in the following chapters for the particle and string),
we perform an analysis similar to that of (3.3.5,6): Making the (almost) unitary
transformation [2.3]  

ln U = (ln p+ ) c + M 3 , (3.4.3a)
where xα = (c, c̃), M αa = (M +a , M −a ), and M α m = (M + m , M − m ), we get
∂ ∂
Q → −ic 12 (pa 2 + M 2 ) + M + i + (M +a pa + M + m M) + x− i ,
∂c ∂c̃
∂ ∂
J3 = c + M 3 − c̃ . (3.4.3b)
∂c ∂c̃
(Cf. (3.2.6a,12a).) As in sect. 3.2, the extra terms in x− and c̃ (analogous to B i and
C i ) can be dropped in the IGL(1) formalism. After dropping such terms, J 3† = 1−J 3 .
(Or we can subtract 2 to make it simply antihermitian. However, we prefer not to,
so that physical states will still have vanishing ghost number.)

Since p+ is a momentum, this redefinition has a funny effect on reality (but not
hermiticity) properties: In particular, c is now a momentum rather than a coordi-
nate (because it has been scaled by p+ , maintaining its hermiticity but making it

imaginary in coordinate space). However, we will avoid changing notation or Fourier
transforming the fields, in order to simplify comparison to the OSp(1,1|2) formalism.
The effect of (3.4.3a) on a field satisfying Φ = ΩΦ* is that it now satisfies

Φ = (−1)c∂/∂c+M ΩΦ* (3.4.4)

due to the i in p+ = i∂ + .

These results can be extended to interacting field theory, and we use Yang-Mills
as an example [2.3]. Lorentz-covariantizing the light-cone result (2.1.7,2.4.11), we find

p− = − 14 F ij F ji + 12 (p+ A− )2 ,

J −α = ix− (pα Ai )(p+ Ai ) + ixα − 14 F ij F ji + 12 (p+ A− )2 − Aα p+ A− ,

A− = − [∇i , p+ Ai } . (3.4.5)
p+ 2

When working in the IGL(1) formalism, it’s extremely useful to introduce a
Lorentz covariant type of second-quantized bracket [3.14]. This bracket can be pos-
tulated independently, or derived by covariantization of the light-cone canonical com-
mutator, plus truncation of the c̃, x+ , and x− coordinates. The latter derivation will
prove useful for the derivation of IGL(1) from OSp(1,1|2). Upon covariantization of
the canonical light-cone commutator (2.4.5), the arguments of the fields and of the
δ-function on the right-hand side are extended accordingly. We now have to truncate.
The truncation of x+ is automatic: Since the original commutator was an equal-time
one, there is no x+ δ-function on the right-hand side, and it therefore suffices to
delete the x+ arguments of the fields. At this stage, in addition to x− dependence,
the fields depend on both c and c̃ and the right-hand side contains both δ-functions.
(This commutator may be useful for OSp approaches to field theory.) We now wish
to eliminate the c̃ dependence. This cannot be done by straightforward truncation,
since expansion of the field in this anticommuting coordinate shows that one cannot
eliminate consistently the fields in the c̃ sector. We therefore proceed formally and
just delete the c̃ argument from the fields and the corresponding δ-function, obtaining
[Φ† (1), Φ(2)]c = − δ(x2− − x1− )δ D (x2 − x1 )δ(c2 − c1 ) , (3.4.6)

which is a bracket with unusual statistics because of the anticommuting δ-function on
the right-hand side. The transformation (3.4.3a) is performed next; its nonunitarity
3.4. From the light cone 41

causes the p+ dependence of (3.4.6) to disappear, enabling one to delete the x−
argument from the fields and the corresponding δ-function to find (using (c∂/∂c)c = c)

[Φ† (1), Φ(2)]c = − 12 δ D (x2 − x1 )δ(c2 − c1 ) . (3.4.7)

This is the covariant bracket. The arguments of the fields are (xa , c), namely, the usual
D bosonic coordinates of covariant theories and the single anticommuting coordinate
of the IGL(1) formalism. The corresponding δ-functions appear on the right-hand
side. (3.4.6,7) are defined for commuting (scalar) fields, but generalize straightfor-
wardly: For example, for Yang-Mills, where Ai includes both commuting (Aa ) and
anticommuting (Aα ) fields, [Ai † , Aj ] has an extra factor of η ij . It might be possible
to define the bracket by a commutator [A, B]c = A ∗ B − B ∗ A. Classically it can be
defined by a Poisson bracket:
† 1 δ δ
[A , B]c = 2 dz A B , (3.4.8)
δΦ(z) δΦ(z)
where z are all the coordinates of Φ (in this case, xa and c). For A = B = Φ, the
result of equation (3.4.7) is reproduced. The above equation implies that the bracket
is a derivation:
[A, BC]c = [A, B]c C + (−1)(A+1)B B[A, C]c , (3.4.9)
where the A’s and B’s in the exponent of the (−1) are 0 if the corresponding quantity
is bosonic and 1 if it’s fermionic. This differs from the usual graded Leibnitz rule by
a (−1)B due to the anticommutativity of the dz in the front of (3.4.8), which also
gives the bracket the opposite of the usual statistics: We can write (−1)[A,B]c =
(−1)A+B+1 to indicate that the bracket of 2 bosonic operators is fermionic, etc., a
direct consequence of the anticommutativity of the total δ-function in (3.4.7). One
can also verify that this bracket satisfies the other properties of a (generalized) Lie
[A, B]c = (−1)AB [B, A]c ,
(−1)A(C+1) [A, [B, C]c ]c + (−1)B(A+1) [B, [C, A]c ]c + (−1)C(B+1) [C, [A, B]c ]c = 0 .
Thus the bracket has the opposite of the usual graded symmetry, being antisymmetric
for objects of odd statistics and symmetric otherwise. This property follows from
the hermiticity condition (3.4.4): (−1)c∂/∂c gives (−1)−(∂/∂c)c = −(−1)c∂/∂c upon
integration by parts, which gives the effect of using an antisymmetric metric. The
Jacobi identity has the same extra signs as in (3.4.9). These properties are sufficient
to perform the manipulations analogous to those used in the light cone.

Before applying this bracket, we make some general considerations concerning
the derivation of interacting IGL(1) from OSp(1,1|2). We start with the original
untransformed generators J 3 and J − c = Q. The first step is to restrict our attention
to just the fields at c̃ = 0. Killing all the fields at linear order in c̃ is consistent
with the transformation laws, since the transformations of the latter fields include
no terms which involve only c̃ = 0 fields. Since the linear-in-c̃ fields are canonically
conjugate to the c̃ = 0 fields, the only terms in the generators which could spoil this
property would themselves have to depend on only c̃ = 0 fields, which, because of the
dc̃ (= ∂/∂c̃) integration, would require explicit c̃-dependence. However, from (2.4.15),
since η cc = C cc = 0, we see that J − c anticommutes with pc , and thus has no explicit
c̃-dependence at either the free or interacting levels. (The only explicit coordinate
dependence in Q is from a c term.)

The procedure of restricting to c̃ = 0 fields can then be implemented very simply
by dropping all pc (= −∂/∂c̃)’s in the generators. As a consequence, we also lose all
explicit x− terms in Q. (This follows from [J − c , p+ ] = −pc .) Since c̃ and ∂/∂c̃ now
occur nowhere explicitly, we can also kill all implicit dependence on c̃: All fields are
evaluated at c̃ = 0, the dc̃ is removed from the integral in the generators, and the
δ(c̃2 − c̃1 ) is removed from the canonical commutator, producing (3.4.6). In the case
of Yang-Mills fields Ai = (Aa , Aα ) = (Aa , Ac , Ac̃ ), the BRST generator at this point
is given by 
Q= ic[− 14 F ij F ji + 12 (p+ A− )2 ] − Ac p+ A− , (3.4.11)

where the integrals are now over just xa , x− , and c, and some of the field strengths
F cc = 2(Ac )2 , other F ic = [∇i , Ac } . (3.4.12)

Before performing the transformations which eliminate p+ dependence, it’s now
convenient to expand the fields over c as

Aa = Aa + cχa ,

Ac = iC + cB ,

Ac̃ = iC + cD , (3.4.13)

where the fields on the right-hand sides are xα -independent. (The i’s have been
chosen in accordance with (3.4.4) to make the final fields real.) We next perform the
dc integration, and then perform as the first transformation the first-quantized one
3.4. From the light cone 43

(3.4.3a), Φ → p+ −J Φ (using the first-quantized J 3 = c∂/∂c + M 3 ), which gives
2 1 2 1
−iQ = 1
4 F ab − 1
2 2B + i[∇a , p+ Aa ] − {C, C̃} − p+ 2 C, C̃
p+ p+ p+
  2 
− 2D + C̃ 2 − 2p+ 2 C̃  C2
+ 2χ − i[∇ , C̃] − 2i p+ A , C̃
a a a
[∇a , C] . (3.4.14)

This transformation also replaces (3.4.6) with (3.4.7), with an extra factor of η ij
for [Ai † , Aj ], but still with the x− δ-function. Expanding the bracket over the c’s,

[χa , Ab ]c = 12 η ab , [D, C]c = 1
2 , [B, C̃]c = − 12 , (3.4.15)

where we have left off all the δ-function factors (now in commuting coordinates only).
Note that, by (3.4.10), all these brackets are symmetric.

We might also define a second-quantized
3 ∂
J = i
A p+ c + M 3 Ai , (3.4.16a)
but this form automatically keeps just the antihermitian part of the first-quantized
operator c∂/∂c = 12 [c, ∂/∂c] + 12 : Doing the c integration and transformation (3.4.3a),

J3 = χa Aa − C̃B − 3CD . (3.4.16b)

As a result, the terms in Q of different orders in the fields have different second-
quantized ghost number. Therefore, we use only the first-quantized ghost operator
(or second-quantize it in functional form).

As can be seen in the above equations, despite the rescaling of the fields by
suitable powers of p+ there remains a fairly complicated dependence on p+ . There is
no explicit x− dependence anywhere but, of course, the fields have x− as an argument.
It would seem that there should be a simple prescription to get rid of the p+ ’s in the
transformations. Setting p+ = constant does not work, since it violates the Leibnitz
rule for derivatives (p+ φ = aφ implies that p+ φ2 = 2aφ2 and not aφ2 ). Even setting
p+ φi = λi φi does not work. An attempt that comes very close is the following: Give
the fields some specific x− dependence in such a way that the p+ factors can be
evaluated and that afterwards such dependence can be canceled between the right-
hand side and left-hand side of the transformations. In the above case it seems that

the only possibility is to set every field proportional to (x− )0 but then it is hard
to define 1/p+ and p+ . One then tries setting each field proportional to (x− ) and
then let
→ 0 at the end. In fact this prescription gives the correct answer for the
quadratic terms of the Yang-Mills BRST transformations. Unfortunately it does not
give the correct cubic terms.

It might be possible to eliminate p+ -dependence simply by applying J −+ = 0 as a
constraint. However, this would require resolving some ambiguities in the evaluation
of the nonlocal (in x− ) operator p+ in the interaction vertices.

We therefore remove the explicit p+ -dependence by use of an explicit transforma-
tion. In the Yang-Mills case, this transformation can be completely determined by
choosing it to be the one which redefines the auxiliary field B in a way which elimi-
nates interaction terms in Q involving it, thus making B + i 12 p · A BRST-invariant.
The resulting transformation [3.14] redefines only the BRST auxiliary fields:

Q → eL∆ Q , LA B ≡ [A, B]c , L∆ 2 = 0 ,
1 1
∆= C̃ i[Aa , p+ Aa ] − C̃ 2 C + 2 C̃ (p+ 2 C) , (3.4.17)
p+ p+
simply redefines the auxiliary fields to absorb the awkward interaction terms in
(3.4.14). (We can also eliminate the free terms added to B and χ by adding a

term C̃ipa Aa to ∆ to make the first term C̃(1/p+ )i[∇a , p+ Aa ].) We then find for
the transformed BRST operator

−iQ = 1 2 
− 12 (2B + ip · A)2 − 2DC 2 + (2χa − ipa C)[∇
4 F ab a , C] . (3.4.18)

The resulting transformations are then

QAa = − i[∇a , C] ,
 + p (B + i 1 p · A) ,
Qχa = i 12 [∇b , Fba ] + i{C, χa − i 12 pa C} a 2

QC = − 2i(B + i 12 p · A) ,

QD = − i ∇a , χa − i 12 pa C + i[C, D] ,

QC = iC 2 ,

QB = − 12 pa [∇a , C] . (3.4.19)

Since all the p+ ’s have been eliminated, we can now drop all x− dependence from the
 B of the usual BRST2
fields, integration, and δ-functions. On the fields Aa , C, C,
3.5. Fermions 45

formalism, this result agrees with the corresponding transformations (3.2.8), where
 of (3.2.11). By working with the second-quantized operator form of Q
this B = 12 B
(and of the redefinition ∆), we have automatically obtained a form which makes QΦ
integrable in Φ, or equivalently makes the vertices which follow from this operator
cyclic in all the fields (or symmetric, if one takes group-theory indices into account).
The significance of this property will be described in the next chapter.

This extended-light-cone form of the OSp(1,1|2) algebra will be used to derive
free gauge-invariant actions in the next chapter. The specific form of the generators
for the case of the free open string will be given in sect. 8.2, and the generalization to
the free closed string in sect. 11.1. A partial analysis of the interacting string along
these lines will be given in sect. 12.1.

3.5. Fermions

These results can be extended to fermions [3.15]. This requires a slight modi-
fication of the formalism, since the Sp(2) representations resulting from the above
analysis for spinors don’t include singlets. This modification is analogous to the addi-
tion of the B∂/∂ C terms to Q in (3.2.6a). We can think of the OSp(1,1|2) generators
of (3.4.2) as “orbital” generators, and add “spin” generators which themselves gen-
erate OSp(1,1|2). In particular, since we are here considering spinors, we choose the
spin generators to be those for the simplest spinor representation, the graded gen-
eralization of a Dirac spinor, whose generators can be expressed in terms of graded
Dirac “matrices”:

{γ̃ A , γ̃ B ] = 2η AB , S AB = 14 [γ̃ A , γ̃ B } , J AB  = J AB + S AB , (3.5.1)

where { , ] is the opposite of [ , }. These γ̃ matrices are not to be confused with
the “ordinary” γ matrices which appear in M ij from the dimensional continuation
of the true spin operators. The γ̃ A , like γ i , are hermitian. (The hermiticity of γ i
in the light-cone formalism follows from (γ i )2 = 1 for each i and the fact that all
states in the light-cone formalism have nonnegative norm, since they’re physical.)
The choice of whether the γ̃’s (and also the graded γ i ’s) commute or anticommute
with other operators (which could be arbitrarily changed by a Klein transformation)
follows from the index structure as usual (bosonic for indices ±, fermionic for α).
(Thus, as usual, the ordinary γ matrices γ a commute with other operators, although
they anticommute with each other.)

In order to put the OSp(1,1|2) generators in a form more similar to (3.4.2), we
need to perform unitary transformations which eliminate the new terms in J −+ and
J +α (while not affecting J αβ , although changing J −α ). In general, the appropriate
transformations J AB  = UJ AB U −1 to eliminate such terms are:

ln U = −(ln p+ )S −+ (3.5.2a)

to first eliminate the S −+ term from J −+ , and then

ln U = S + α pα (3.5.2b)

to do the same for J +α . The general result is

J −+ = −ix− p+ + k , J +α = ixα p+ , 
J αβ = −ix(α pβ) + M ,

1  β p + kp + Q

J −α = −ix− pα + −ixα 12 (pb pb + M 2 + pβ pβ ) + M α β α α ; (3.5.3a)
M αβ αβ + S αβ ,

 = (M b p + M M) + S
Q 1 b 2
α α b αm −α + S +α 2 (p pb + M ) . (3.5.3b)

(3.5.3a) is the same as (3.4.2a), but with M αβ and Qα replaced by M 
αβ and Qα . In
this case the last term in Q α

S −α + S +α 12 (pb pb + M 2 ) = − 12 γ̃ α γ̃ − + γ̃ + 21 (pb pb + M 2 ) . (3.5.4)

We can again choose k = 0.

This algebra will be used to derive free gauge-invariant actions for fermions in
sect. 4.5. The generalization to fermionic strings follows from the representation of
the Poincaré algebra given in sect. 7.2.

3.6. More dimensions

In the previous section we saw that fermions could be treated in a way similar to
bosons by including an OSp(1,1|2) Clifford algebra. In the case of the Dirac spinor,
there is already an OSp(D−1,1|2) Clifford algebra (or OSp(D,1|2) in the massive
case) obtained by adding 2+2 dimensions to the light-cone γ-matrices, in terms of
which M ij (and therefore the OSp(1,1|2) algebra) is defined. Including the addi-
tional γ-matrices makes the spinor a representation of an OSp(D,2|4) Clifford algebra
(OSp(D+1,2|4) for massive), and is thus equivalent to adding 4+4 dimensions to the
3.6. More dimensions 47

original light-cone spinor instead of 2+2, ignoring the extra spacetime coordinates.
This suggests another way of treating fermions which allows bosons to be treated
identically, and should thus allow a straightforward generalization to supersymmetric
theories [3.16].

We proceed similarly to the 2+2 case: Begin by adding 4+4 dimensions to the
light-cone Poincaré algebra (2.3.5). Truncate the resulting IOSp(D+1,3|4) algebra to
ISO(D−1,1)⊗IOSp(2,2|4). IOSp(2,2|4) contains (in particular) 2 inequivalent trun-
cations to IOSp(1,1|2), which can be described by (defining-representation direct-
product) factorization of the OSp(2,2|4) metric into the OSp(1,1|2) metric times the
metric of either SO(2) (U(1)) or SO(1,1) (GL(1)):

η AB = η AB η âb̂ , A = Aâ

→ J AB = η b̂â J Aâ,Bb̂ ,
âb̂ ∆ = η BA J Aâ,Bb̂ , (3.6.1)

where ∆ is the generator of the U(1) or GL(1) and η âb̂ = I or σ 1 . These 2 OSp(1,1|2)’s
are Wick rotations of each other. We’ll treat the 2 cases separately.

The GL(1) case corresponds to first taking the GL(2|2) (=SL(2|2)⊗GL(1)) sub-
group of OSp(2,2|4) (as SU(N)⊂SO(2N), or GL(1|1)⊂OSp(1,1|2)), keeping also half
of the inhomogeneous generators to get IGL(2|2). Then taking the OSp(1,1|2) sub-
group of the SL(2|2) (in the same way as SO(N)⊂SU(N)), we get IOSp(1,1|2)⊗GL(1),
which is like the Poincaré group in (1,1|2) dimensions plus dilatations. (There is also
an SL(1|2)=OSp(1,1|2) subgroup of SL(2|2), but this turns out not to be useful.) The
advantage of breaking down to GL(2|2) is that for this subgroup the coordinates of
the string (sect. 8.3) can be redefined in such a way that the extra zero-modes are
separated out in a natural way while leaving the generators local in σ. This GL(2|2)
subgroup can be described by writing the OSp(2,2|4) metric as
0 η AB
η AB = , η AB = (−1)AB η B A , A = (A, A ) (3.6.2a)
η A B 0
→ J AB = , J˜AB = −(−1)AB J˜B A , (3.6.2b)
where J˜A B are the GL(2|2) generators, to which we add the p̃A half of pA to form
IGL(2|2). (The metric η AB can be used to eliminate primed indices, leaving covariant
and contravariant unprimed indices.) In this notation, the original ± indices of the
light cone are now + and − (whereas + and − are “transverse”). To reduce to

the IOSp(1,1|2)⊗GL(1) subgroup we identify primed and unprimed indices; i.e., we
choose the subgroup which transforms them in the same way:

J AB = J˜A B + J˜AB , p̌A = p̃A , ∆ = η BA J˜A B . (3.6.3)

We distinguish the momenta p̌A and their conjugate coordinates x̌A , which we wish
to eliminate, from pA = p̃A and their conjugates xA , which we’ll keep as the usual
ones of OSp(1,1|2) (including the nonlinear p− ). At this point these generators take
the explicit form

∆ = ix̌A p̌A − ix− p+ − ixα pα + M − + + C βα M α β ,

J +α = ip̌+ x̌α − ix̌+ p̌α + ip+ xα + M +α ,

J −+ = −ix̌− p̌+ + ix̌+ p̌− − ix− p+ + M − + ,

J αβ = −ix̌(α p̌β) − ix(α pβ) + M αβ  + M α β ,
J −α = −ix̌− p̌α + ix̌α p̌− − ix− pα + ixα p− + M − α + Qα ,
p− = − (pa 2 + M 2 + 2p+ p̌− + 2pα p̌α ) ,
Qα = M α a pa + M α m M + M α + p̌− + M α − p+ − M α β p̌β − M α β  pβ . (3.6.4)

(All the p̌’s are linear, being unconstrained so far.) We now use p and ∆ to eliminate
the extra zero-modes. We apply the constraints and corresponding gauge conditions
∆=0 → x̌− = (x− p+ + xα pα − x̌+ p̌− − x̌α p̌α + iM − + + iC βα M α β ) ,
gauge p̌+ = 1 ;

p̌A p̌A = 0 → p̌− = − 12 p̌α p̌α ,

gauge x̌+ = 0 . (3.6.5)

These constraints are directly analogous to (2.2.3), which were used to obtain the
usual coordinate representation of the conformal group SO(D,2) from the usual co-
ordinate representation (with 2 more coordinates) of the same group as a Lorentz
group. In fact, after making a unitary transformation of the type (3.5.2b),

U = e−(ip+ xα +M +α )p̌
, (3.6.6)

the remaining unwanted coordinates x̌α completely decouple:

UJ AB U −1 = J AB + J  AB , (3.6.7a)
3.6. More dimensions 49

◦ ◦ ◦ ◦
J +α = ix̌α , J −+ = ix̌α p̌α , J αβ = −ix̌(α p̌β) , J −α = −ix̌α p̌β p̌β ;
  α βα
J +α = 2(ip+ xα + M +α ) , J −+ = 2(−ix− p+ + M − + ) + (−ix pα + C M α β ) ,

J  αβ = −ix(α pβ) + (M α β + M αβ  ) ,

J  −α = (−ix− pα + M − α )

+ −ixα 21 (pa 2 + M 2 ) + (M α a pa + M α m M + M α − p+ − M α β  pβ ) ,


where J are the generators of the conformal group in 2 anticommuting dimensions
((3.3.9), after switching coordinates and momenta), and J  are the desired OSp(1,1|2)
To eliminate zero-modes, it’s convenient to transform these OSp(1,1|2) generators
to the canonical form (3.4.2a). This is performed [3.7] by the redefinition

1 2
p+ → 2 p+ , (3.6.8a)

followed by the unitary transformations

− −i 2 [xα ,pα ]+2M − + +C βα M α β
U 1 = p+ ,

U 2 = e−2M +α p
. (3.6.8b)

Since p+ is imaginary (though hermitian) in (x− ) coordinate space, U 1 changes reality
conditions accordingly (an i for each p+ ). The generators are then

J +α = ixα p+ , J −+ = −ix− p+ , 
J αβ = −ix(α pβ) + M ,

1  βp + Q

J −α = −ix− pα + −ixα 12 (pa 2 + M 2 + pβ pβ ) + M α β α ; (3.6.9a)
 =M  +M 
M ,
αβ αβ αβ

 = M  − 1 M   + (M  a p + M  M) + M  (p 2 + M 2 ) .
Q (3.6.9b)
α −α 2 −α α a αm +α a

For the U(1) case the derivation is a little more straightforward. It corresponds
to first taking the U(1,1|1,1) (=SU(1,1|1,1)⊗ U(1)) subgroup of OSp(2,2|4). From
(3.6.1), instead of (3.6.2a,3) we now have
η AB 0
η AB = , η AB = η A B , A = (A, A ) ,
0 η A B 

J AB = J˜AB + J˜A B , p̌A = p̃A , ∆ = η BA J˜A B . (3.6.10)
The original light-cone ± are now still ± (no primes), so the unwanted zero-modes
can be eliminated by the constraints and gauge choices

p̌A = 0 → gauge x̌A = 0 . (3.6.11)

Alternatively, we could include p̌A among the generators, using IOSp(1,1|2) as the
group (as for the usual closed string: see sects. 7.1, 11.1). (The same result can be
obtained by replacing (3.6.11) with the constraints ∆ = 0 and
b̂â pAâ pBb̂ ∼ p[A p̌B) =
0.) The OSp(1,1|2) generators are now

J +α = ixα p+ + M + α , J −+ = −ix− p+ + M − + ,

J αβ = −ix(α pβ) + M αβ + M α β  ,
1 1
J −α = −ix− pα −ixα (pa 2 +M 2 +pβ pβ )+ (M α a pa +M αm M +M α β pβ )+M − α ,
2p+ p+
or, in other words (symbols), these OSp(1,1|2) generators are just the usual ones plus
the spin of a second OSp(1,1|2), with the same representation as the spin of the first
J AB = J˜AB + M A B . (3.6.12b)
(However, for the string M αm M will contain oscillators from both sets of 2+2 dimen-
sions, so these sets of oscillators won’t decouple, even though J˜AB commutes with
M A B .) To simplify the form of J −+ and J −α , we make the consecutive unitary
transformations (3.5.2):
α p
U 1 = p+ −M − + , U 2 = eM + α
, (3.6.13)

after which the generators again take the canonical form:

J +α = ixα p+ , J −+ = −ix− p+ 
, J αβ = −ix(α pβ) + M ,

1  βp + Q

J −α = −ix− pα + −ixα 21 (pa 2 + M 2 + pβ pβ ) + M α β α ; (3.6.14a)
M αβ αβ + M α β  ,
 = M   + (M b p + M M) + M   1 (p 2 + M 2 ) .
Q (3.6.14b)
α −α α b αm +α 2 a

Because of U 1 , formerly real fields now satisfy φ† = (−1)M − + φ.
Examples and actions of this 4+4-extended OSp(1,1|2) will be considered in sect.
4.1, its application to supersymmetry in sect. 5.5, and its application to strings in
sect. 8.3.
Exercises 51


(1) Derive the time derivative of (3.1.6) from (3.1.7).

(2) Derive (3.1.12). Compare with the usual derivation of the Noether current in
field theory.

(3) Derive (3.1.15,17).

(4) Show that Q of (3.2.6a) is nilpotent. Show this directly for (3.2.8).

(5) Derive (3.3.1b).

(6) Use (3.3.6) to rederive (3.2.6a,12a).

(7) Use (3.3.2,7) to derive the OSp(1,1|2) algebra for Yang-Mills in terms of the
explicit independent fields (in analogy to (3.2.8)).

(8) Perform the transformation (3.4.3a) to obtain (3.4.3b). Choose the Dirac spinor
representation of the spin operators (in terms of γ-matrices). Compare with
(3.2.6), and identify the field equations G and ghosts C.
 and M
(9) Check that the algebra of Q 
α αβ closes for (3.5.3), (3.6.9), and (3.6.14),

and compare with (3.4.2).


4.1. OSp(1,1|2)

In this chapter we will use the results of sects. 3.4-6 to derive free gauge-invariant
actions for arbitrary field theories, and discuss some preliminary results for the ex-
tension to interacting theories.

The (free) gauge covariant theory for arbitrary representations of the Poincaré
group (except perhaps for those satisfying self-duality conditions) can be constructed
from the BRST1 OSp(1,1|2) generators [2.3]. For the fields described in sect. 3.4
which are representations of OSp(D,2|2), consider the gauge invariance generated by
OSp(1,1|2) and the obvious (but unusual) corresponding gauge-invariant action:

δΦ = 1 BA
2J ΛAB → S= dD xa dx− d2 xα 21 Φ† p+ δ(J AB )Φ , (4.1.1)

where J AB for A = (+, −, α) (graded antisymmetric in its indices) are the generators
of OSp(1,1|2), and we have set k = 0, so that the p+ factor is the Hilbert space
metric. In particular, the J −+ and J +α transformations allow all dependence on the
unphysical coordinates to be gauged away:

δΦ = −ix− p+ Λ−+ + ixα p+ Λ−α (4.1.2)

implies that only the part of Φ at x− = xα = 0 can be gauge invariant. A more
explicit form of δ(J AB ) is given by

p+ δ(J AB ) = p+ δ(J αβ 2 )iδ(J −+ )δ 2 (J +α )δ 2 (J −α )

= δ(x− )δ 2 (xα )δ(M αβ 2 )p+ 2 J −α 2 , (4.1.3)

where we have used

J −+ δ(J −+ ) = δ(J −+ )J −+ = 0 → δ(J −+ ) = i δ(x− ) , (4.1.4)
4.1. OSp(1,1|2) 53

since p+ = 0 in light-cone formalisms. The gauge invariance of the kinetic operator
follows from the fact that the δ-functions can be reordered fairly freely: δ(J αβ 2 )
(which is really a Kronecker δ) commutes with all the others, while

δ(J −+ + a)δ 2 (J ±α ) = δ 2 (J ±α )δ(J −+ + a ∓ 2) → [δ(J −+ ), δ 2 (J +α )δ 2 (J −α )] = 0 ,

[δ 2 (J −α ), δ 2 (J +β )] = 2J −+ + (C αβ J −+ + J αβ ) 12 [J −α , J +β ] , (4.1.5)

where the J −+ and J αβ each can be freely moved to either side of the [J −α , J +β ].
After integration of the action over the trivial coordinate dependence on x− and xα ,
(4.1.1) reduces to (using (3.4.2,4.1.3))

S= dD xa 12 φ† δ(M αβ 2 )(2 −M 2 +Q2 )φ , δφ = −i 12 Qα Λα + 12 M αβ Λαβ , (4.1.6)

where φ now depends only on the usual spacetime coordinates xa , and for irreducible
Poincaré representations φ has indices which are the result of starting with an irre-
ducible representation of OSp(D−1,1|2) in the massless case, or OSp(D,1|2) in the
massive case, and then truncating to the Sp(2) singlets. (This type of action was first
proposed for the string [4.1,2].) Λα is the remaining part of the J −α transformations
after using up the transformations of (4.1.2) (and absorbing a 1/∂ + ), and contains
the usual component gauge transformations, while Λαβ just gauges away the Sp(2)
nonsinglets. We have thus derived a general gauge-covariant action by adding 2+2
dimensions to the light-cone theory. In sect. 4.4 we’ll show that gauge-fixing to the
light cone gives back the original light-cone theory, proving the consistency of this

In the BRST formalism the field contains not only physical polarizations, but
also auxiliary fields (nonpropagating fields needed to make the action local, such as
the trace of the metric tensor for the graviton), ghosts (including antighosts, ghosts
of ghosts, etc.), and Stueckelberg fields (gauge degrees of freedom, such as the gauge
part of Higgs fields, which allow more renormalizable and less singular formalisms for
massive fields). All of these but the ghosts appear in the gauge-invariant action. For
example, for a massless vector we start with Ai = (Aa , Aα ), which appears in the field
φ as
  i j
φ = i Ai ,  = η ij . (4.1.7)

Reducing to Sp(2) singlets, we can truncate to just Aa . Using the relations
M ij k = [i η j)k → M αa b = η ab α , M αa β = −C αβ a , (4.1.8)

where [ ) is graded antisymmetrization, we find
c a
1 αb
2M Mα  = 12 M αb η ac α = η ac b , (4.1.9)

and thus the lagrangian
L= 1
2 φδ(M αβ 2 )[2 − (M α b ∂ b )2 ]φ = 12 Aa (2 Aa − ∂ a ∂ b Ab ) . (4.1.10)

Similarly, for the gauge transformation
Λ = i Λi α → δAa = 1 b  α
a  2 M α ∂ b Λ = − 12 ∂ a Λα α . (4.1.11)

As a result of the δ(M αβ 2 ) acting on Qα Λα , the only part of Λ which survives is
the part which is an overall singlet in the matrix indices and explicit α index: in
this case, Λi α = δ i α λ → δAa = −∂ a λ. Note that the φQ2 φ term can be written as
a (Qφ)2 term: This corresponds to subtracting out a “gauge-fixing” term from the
“gauge-fixed” lagrangian φ(2 − M 2 )φ. (See the discussion of gauge fixing in sect.

For a massless antisymmetric tensor we start with A[ij) = (A[ab] , Aaα , A(αβ) ) ap-
pearing as
φ = − 12 ij = √ [i ⊗ j)
Aji ,  (4.1.12)
A A 2
(and similarly for |Λα
), and truncate to just A[ab] . Then, from (4.1.8),

1 d a b    
2M Mα   = 12 M αc η da α b + η db a α
= η da c b + η db a c + η d(a η b)c 12 α α , (4.1.13)

and we have

L = 14 Aab (2 Aab + ∂ c ∂ [a Ab]c ) , δAab = 12 ∂ [a Λb]α α . (4.1.14)

For a massless traceless symmetric tensor we start with h(ij] = (h(ab) , hαb , h[αβ] )
satisfying hi i = ha a + hα α = 0, appearing as
φ = 12 ij hji ,  = √ (i j] , (4.1.15)
S S 2
1 ab
and truncate to (h(ab) , h[αβ] ), where h[αβ] = 2 C αβ η h(ab) , leaving just an uncon-
strained symmetric tensor. Then, using (4.1.13), as well as
b α β
1 γa
2M Mγ   = 12 C αβ (a b) − η ab α β , (4.1.16)
4.1. OSp(1,1|2) 55

and using the condition hα α = −ha a , we find

L = 14 hab 2 hab − 12 hab ∂ b ∂ c ha c + 12 hc c ∂ a ∂ b hab − 14 ha a 2 hb b ,

δhab = − 12 ∂ (a Λb)α α . (4.1.17)
This is the linearized Einstein-Hilbert action for gravity.
The massive cases can be obtained by the dimensional reduction technique, as
in (2.2.9), since that’s how it was done for this entire procedure, from the light-cone
Poincaré algebra down to (4.1.6). (For the string, the OSp generators are represented
in terms of harmonic oscillators, and M αm M is cubic in those oscillators instead of
quadratic, so the oscillator expressions for the generators don’t follow from dimen-
sional reduction, and (4.1.6) must be used directly with the M αm M terms.) Techni-
cally, pm = m makes sense only for complex fields. However, at least for free theories,
the resulting i’s that appear in the pa pm crossterms can be removed by appropriate
redefinitions for the complex fields, after which they can be chosen real. (See the dis-
cussion below (2.2.10).) For example, for the massive vector we replace Am → iAm
(and then take Am real) to obtain
φ = a Aa + im Am + α Aα , φ = Aa a  − iAm m + A
α . (4.1.18)

The lagrangian and invariance then become

L = 12 Aa [(2 − m2 )Aa − ∂ a ∂ b Ab ] + 12 Am 2 Am + mAm ∂ a Aa

= 14 F ab 2 − 12 (mAa + ∂ a Am )2 ,

δAa = −∂ a λ , δAm = mλ . (4.1.19)
This gives a Stueckelberg formalism for a massive vector.
Other examples reproduce all the special cases of higher-spin fields proposed ear-
lier [4.3] (as well as cases that hadn’t been obtained previously). For example, for
totally symmetric tensors, the usual “double-tracelessness” condition is automatic:
Starting from the light cone with a totally symmetric and traceless tensor (in trans-
verse indices), extending i → (a, α) and restricting to Sp(2) singlets, directly gives a
totally symmetric and traceless tensor (in D-dimensional indices) of the same rank,
and one of rank 2 lower (but no lower than that, due to the total antisymmetry in
the Sp(2) indices).
The most important feature of the BRST method of deriving gauge-invariant
actions from light-cone (unitary) representations of the Poincaré group is that it

automatically includes exactly the right number of auxiliary fields to make the action
local. In the case of Yang-Mills, the auxiliary field (A− ) was obvious, since it results
directly from adding just 2 commuting dimensions (and not 2 anticommuting) to the
light cone, i.e., from making D − 2-dimensional indices D-dimensional. Furthermore,
the necessity of this field for locality doesn’t occur until interactions are included
(see sect. 2.1). A less trivial example is the graviton: Naively, a traceless symmetric
D-dimensional tensor would be enough, since this would automatically include the
analog of A− . However, the BRST method automatically includes the trace of this
tensor. In general, the extra auxiliary fields with anticommuting “ghost-valued”
Lorentz indices are necessary for gauge-covariant, local formulations of field theories
[4.4,5]. In order to study this phenomenon in more detail, and because the discussion
will be useful later in the 2D case for strings, we now give a brief discussion of general
General relativity is the gauge theory of the Poincaré group. Since local trans-
lations (i.e., general coordinate transformations) include the orbital part of Lorentz
transformations (as translation by an amount linear in x), we choose as the group
generators ∂ m and the Lorentz spin M ab . Treating M ab as second-quantized opera-
tors, we indicate how they act by writing explicit “spin” vector indices a, b, . . . (or
spinor indices) on the fields, while using m, n, . . . for “orbital” vector indices on which
M ab doesn’t act, as on ∂ m . (The action of the second-quantized M ab follows from that
of the first-quantized: E.g., from (4.1.8), (2.2.5), and the fact that (M ij )† = −M ij ,
we have M ab Ac = −η c[a Ab] .) The spin indices (but not the orbital ones) can be con-
tracted with the usual constant tensors of the Lorentz group (the Lorentz metric and
γ matrices). The (antihermitian) generators of gauge transformations are thus

λ = λm (x)∂ m + 12 λab (x)M ba , (4.1.20)

and the covariant derivatives are

∇a = ea m ∂ m + 12 ω a bc M cb , (4.1.21)

where we have absorbed the usual derivative term, since derivatives are themselves
generators, and to make the covariant derivative transform covariantly under the
gauge transformations
∇a  = eλ ∇a e−λ . (4.1.22)
Covariant field strengths are defined, as usual, by commutators of covariant deriva-
[∇a , ∇b ] = T ab c ∇c + 12 Rab cd M dc , (4.1.23)
4.1. OSp(1,1|2) 57

since that automatically makes them transform covariantly (i.e., by a similarity trans-
formation, as in (4.1.22)), as a consequence of the transformation law (4.1.22) of the
covariant derivatives themselves. Without loss of generality, we can choose

T ab c = 0 , (4.1.24)

since this just determines ω ab c in terms of ea m , and any other ω can always be written
as this ω plus a tensor that is a function of just T . (The theory could then always be
rewritten in terms of the T = 0 ∇ and T itself, making T an arbitrary extra tensor
with no special geometric significance.) To solve this constraint we first define

ea = ea m ∂ m ,

[ea , eb ] = cab c ec . (4.1.25)

cab c can then be expressed in terms of ea m , the matrix inverse em a ,

ea m em b = δ a b , em a ea n = δ m n , (4.1.26)

and their derivatives. The solution to (4.1.24) is then

ω abc = 12 (cbca − ca[bc] ) . (4.1.27)

The usual global Lorentz transformations, which include orbital and spin pieces
together in a specific way, are a symmetry of the vacuum, defined by

= κ∂ a ↔ ea m
= κδ a m . (4.1.28)

κ is an arbitrary constant, which we can choose to be a unit of length, so that
∇ is dimensionless. (In D = 4 it’s just the usual gravitational coupling constant,
proportional to the square root of Newton’s gravitational constant.) As a result of
general coordinate invariance, any covariant object (i.e., a covariant derivative or
tensor with only spin indices uncontracted) will then also be dimensionless. The
subgroup of the original gauge group which leaves the vacuum (4.1.28) invariant
is just the usual (global) Poincaré group, which treats orbital and spin indices in
the same way. We can also treat these indices in a similar way with respect to
the full gauge group by using the “vielbein” ea m and its inverse to convert between
spin and orbital indices. In particular, the orbital indices on all fields except the
vielbein itself can be converted into spin ones. Also, since integration measures are
antisymmetric, converting dxm into Ωa = dxm em a converts dD x into ΩD = dD x e−1 ,

where e = det(ea m ). On such covariant fields, ∇ always acts covariantly. On the
other hand, in the absence of spinors, all indices can be converted into orbital ones.
In particular, instead of the vielbein we could work with the metric tensor and its
g mn = η ab em a en b , g mn = η ab ea m eb n . (4.1.29)

Then, instead of ∇, we would need a covariant derivative which knows how to treat
uncontracted orbital indices covariantly.

The action for gravity can be written as

S = − 12 dD x e−1 R , R = 12 Rab ba . (4.1.30)

This can be rewritten in terms of cabc as

e−1 R = −∂ m (e−1 eam ca b b ) + e−1 − 12 (cab b )2 + 18 cabc cabc − 14 cabc cbca (4.1.31)

e−1 ea f a = ∂ m (e−1 eam f a ) + e−1 cab b f a . (4.1.32)

Expanding about the vacuum,

ea m = κδ a m + κD/2 ha m , (4.1.33)

where we can choose eam (and thus ham ) to be symmetric by the λab transformation,
the linearized action is just (4.1.17). As an alternative form for the action, we can
consider making the field redefinition

ea m → φ−2/(D−2) ea m , (4.1.34)

which introduces the new gauge invariance of (Weyl) local scale transformations

ea m = eζ ea m , φ = e(D−2)ζ/2 φ . (4.1.35)

(The gauge choice φ = constant returns the original fields.) Under the field redefini-
tion (4.1.34), the action (4.1.30) becomes

S→ dD x e−1 2 D−1 2 1
D−2 (∇a φ) − 2 Rφ
. (4.1.36)

We have actually started from (4.1.30) without the total-derivative term of (4.1.31),
which is then a function of just ea m and its first derivatives, and thus correct even at
boundaries. (We also dropped a total-derivative term −∂ m ( 12 φ2 e−1 eam ca b b ) in (4.1.36),
4.1. OSp(1,1|2) 59

which will be irrelevant for the following discussion.) If we eliminate φ by its field
equation, but keep surface terms, this becomes

S → dD x e−1 2 D−1
D−2 ∇ · (φ∇φ)

= dD x e−1 D−1
D−2 2 φ2

= dD x e−1 D−1
D−2 2 φ
2 + 2 φ
(φ − φ
) + (φ − φ
)2 (4.1.37)

We can solve the φ field equation as
φ = φ
1 − R . (4.1.38)
4 D−1
2 +R
(We can choose φ
= 1, or take the κ out of (4.1.28) and introduce it instead through
= κ−(D−2)/2 by a global ζ transformation.) Assuming φ falls off to φ
fast enough
at ∞, the last term in (4.1.37) can be dropped, and, using (4.1.38), the action becomes

−1 1
S→ − 12 D
d xe R−R R . (4.1.39)
4 D−1
2 +R
Since this action has the invariance (4.1.35), we can gauge away the trace of h or,
equivalently, gauge the determinant of ea m to 1. In fact, the same action results from
(4.1.30) if we eliminate this determinant by its equation of motion.
Thus, we see that, although gauge-covariant, Lorentz-covariant formulations are
possible without the extra auxiliary fields, they are nonlocal. Furthermore, the non-
localities become more complicated when coupling to nonconformal matter (such as
massive fields), in a way reminiscent of Coulomb terms or the nonlocalities in light-
cone gauges. Thus, the construction of actions in such a formalism is not straightfor-
ward, and requires the use of Weyl invariance in a way analogous to the use of Lorentz
invariance in light-cone gauges. Another alternative is to eliminate the trace of the
metric from the Einstein action by a coordinate choice, but the remaining constrained
(volume-preserving) coordinate invariance causes difficulties in quantization [4.7].
We have also seen that some properties of gravity (the ones relating to conformal
transformations) become more transparent when the scale compensator φ is intro-
duced. (This is particularly true for supergravity.) Introducing such fields into the
OSp formalism requires introducing new degrees of freedom, to make the representa-
tion larger (at least in terms of gauge degrees of freedom). Although such invariances
are hard to recognize at the free level, the extensions of sect. 3.6 show signs of per-
forming such generalizations. However, while the U(1)-type extension can be applied

to arbitrary Poincaré representations, the GL(1)-type has difficulty with fermions.
We’ll first discuss this difficulty, then show how the 2 types differ for bosons even for
the vector, and finally look again at gravity.

The U(1) case of spin 1/2 reproduces the algebra of sect. 3.5, since M A B of
(3.6.12b) is exactly the extra term of (3.5.1):

M ij = 12 γ ij →

 2 )(2 + Q
φ̄δ(M  2 )φ = φ̄δ( )eiγ + p//2 i 1 p iγ + p//2
αβ 2 /γ − e φ

= i 14 (γ − φ + i 12 γ − γ + p/φ)δ( )γ + p/(γ − φ + i 12 γ − γ + p/φ)
ˆ p/γ  φ̂ ,
= i 12 φ̄ (4.1.40)

where γ ij = 12 [γ i , γ j }, and we have used
0 = 18 (γ αβ + γ α β )(γ αβ + γ α β  ) = (γ α γ α )2 + 4 → γ α γ α = 2i . (4.1.41)

(We could equally well have chosen the other sign. This choice, with our conven-
tions, corresponds to harmonic-oscillator boundary conditions: See sect. 4.5.) After
eliminating γ + φ by gauge choice or, equivalently, by absorbing it into γ − φ by field re-
definition, this becomes just ϕ̄/
pϕ. However, in the GL(1) case, the analog to (4.1.41)
0 = 18 ({γ α , γ β } + {γ α , γ β })({γ α , γ β  } + {γ α , γ β }) = (γ α γ α )(γ β γ β ) ,

γ α γ α + γ α γ α = −4 , (4.1.42)

and to (4.1.40) is

2 + Q 2 = −p2 + 18 (γ α γ α )γ − (/p − γ + p2 ) + 18 (γ α γ α )(/p − γ + p2 )γ − .


Unfortunately, φ and φ̄ must have opposite boundary conditions γ α γ α = 0 or γ α γ α =
 2 ), as is evidenced by the asymmetric
0 in order to contribute in the presence of δ(M αβ

form of (4.1.43) for either choice. Consequently, the parts of φ and φ̄ that survive
are not hermitian conjugates of each other, and the action is not unitary. (Properly
speaking, if we choose consistent boundary conditions for both φ and φ̄, the action
vanishes.) Thus, the GL(1)-type OSp(1,1|2) is unsuitable for spinors unless further
modified. In any case, such a modification would not treat bosons and fermions
symmetrically, which is necessary for treating supersymmetry. (Fermions in the usual
OSp formalism will be discussed in more detail in sect. 4.5.)
4.1. OSp(1,1|2) 61

For the case of spin 1 (generalizing the light-cone Hilbert space, as in (4.1.7-8)),
we expand

φ = |a
Aa + i |−
A+ + i |+
A− , φ† = Aa a | − iA+ − | − iA− + | , (4.1.44)

for the GL(1) case, and the same for U(1) with |+
→ |+
(Sp(2)-spinor fields again
drop out of the full φ). We find for GL(1) [3.7]

L = − 14 F ab 2 − 12 (A− + ∂ · A)2 = − 14 F ab 2 − 12 Â− 2 , (4.1.45a)

where F ab = ∂ [a Ab] , and for U(1)

L = − 14 F ab 2 + 12 (A− + 12 2 A+ )2 = − 14 F ab 2 + 12 Â− 2 . (4.1.45b)

In both cases A+ can be gauged away, and A− is auxiliary. However, the sign for
U(1)-type OSp(1,1|2) is the same as for auxiliary fields in supersymmetry (for off-shell
irreducible multiplets), whereas the sign for GL(1) is opposite. The sign difference
is not surprising, considering the U(1) and GL(1) types are Wick rotations of each
other: This auxiliary-field term, together with the auxiliary component of Aa (the
light-cone A− ), appear with the metric η âb̂ of (3.6.1), and thus with the same sign
for SO(2) (U(1)). In fact, (4.1.45b) is just the part of the 4D N=1 super-Yang-Mills
lagrangian for fields which are R-symmetry invariant: A− can be identified with the
usual auxiliary field, and A+ with the θ = 0 component of the superfield. Similarly,
γ + φ for spin 1/2 can be identified with the linear-in-θ part of this superfield. This
close analogy strongly suggests that the nonminimal fields of this formalism may
be necessary for treating supersymmetry. Note also that for GL(1) the auxiliary
automatically mixes with the spin-1 “gauge-fixing” function, like a Nakanishi-Lautrup
field, while for U(1) there is a kind of “parity” symmetry of the OSp(1,1|2) generators,
→ − |A
, which prevents such mixing, and can be included in the usual parity
transformations to strengthen the identification with supersymmetry.
For spin 2, for U(1) we define
1     1     1    
φ = 12 hab √ (a b) + iAa + √ (a − ) + iAa − √ (a + )
2 2 2

+ ϕ++ |−
+ ϕ+− √ (+ − ) + ϕ−− |+

+ ϕ √ |α

+ ϕ 12 α α + α α + ϕ √ α α , (4.1.46a)
2 2
subject to the tracelessness condition (hi i = 0)
1 a
2h a + ϕ+− − ϕ − ϕ = 0 , (4.1.46b)

and find the lagrangian

L = − 14 hab (p2 hab − 2pa pc hcb + 2pa pb hc c − η ab p2 hc c )

+ 12 (Aa− − 12 p2 Aa+ + pa pb Ab+ − i 2pa ϕ )2
√ 1
+ (ϕ+− − ϕ )( 2ϕ−− + p2 ϕ+− − 2p2 ϕ + √ p4 ϕ++ + p2 hb b − pb pc hbc )
2 2
1 2

= “R” + 2 Âa− + 2ϕ̂+− ϕ̂−− . (4.1.47)

The second term is the square of an auxiliary “axial” vector (which again appears
with sign opposite to that in GL(1) [3.7]), which resembles the axial vector auxiliary
field of supergravity (including terms which can be absorbed, as for spin 1). In the
last term, the redefinition ϕ−− → ϕ̂−− involves the (linearized) Ricci scalar. Although
it’s difficult to tell from the free theory, it may also be possible to identify some of
the gauge degrees of freedom with conformal compensators: ϕ with the compensator
for local R-symmetry, and ϕ+− (or ϕ or ϕ ; one is eliminated by the tracelessness
condition and one is auxiliary) with the local scale compensator.
A simple expression for interacting actions in terms of just the OSp(1,1|2) group
generators has not yet been found. (However, this is not the case for IGL(1): See the
following section.) The usual gauge-invariant interacting field equations can be de-
rived by imposing J αβ φ = J −α φ = 0, which are required in a (anti)BRST formalism,
and finding the equations satisfied by the x− = xα = 0 sector. However, this requires
use of the other sectors as auxiliary fields, whereas in the approach described here
they would be gauge degrees of freedom.
These results for gauge-invariant actions from OSp(1,1|2) will be applied to the
special case of the string in chapt. 11.

4.2. IGL(1)
We now derive the corresponding gauge-invariant action in the IGL(1) formalism
and compare with the OSp(1,1|2) results. We begin with the form of the generators
(3.4.3b) obtained from the transformation (3.4.3a). For the IGL(1) formalism we can
then drop the zero-modes x− and c̃, and the action and invariance then are (using
δ(Q) = Q)

S=− dD xdc Φ† iQδ(J 3 ) Φ , δΦ = −iQΛ + J 3 Λ̂ . (4.2.1)

This is the IGL(1) analog of (4.1.1). (This action also was first proposed for the
string [4.8,9].) The δ(J 3 ) kills the sign factor in (3.4.4). However, even though some
4.2. IGL(1) 63

unphysical coordinates have been eliminated, the field is still a representation of the
spin group OSp(D-1,1|2) (or OSp(D,1|2)), and thus there is still a “hidden” Sp(2)
symmetry broken by this action (but only by auxiliary fields: see below).

To obtain the analog of (4.1.6), we first expand the field in the single ghost
zero-mode c:
Φ = φ + icψ . (4.2.2)

φ is the field of the OSp(1,1|2) formalism after elimination of all its gauge zero-
modes, and ψ is an auxiliary field (identified with the Nakanishi-Lautrup auxiliary
fields in the gauge-fixed formalism [4.4,5]). If we expand the action (4.2.1) in c, using
(3.4.3b), and the reality condition on the field to combine crossterms, we obtain, with
Qα = (Q+ , Q− ),

L =− dc Φ† iQδ(J 3 )Φ

= 12 φ† (2 − M 2 )δ(M 3 )φ − ψ † M + δ(M 3 + 1)ψ + 2ψ † Q+ δ(M 3 )φ . (4.2.3)

As an example of this action, we again consider a massless vector. In analogy to
Φ = i φi + ic i ψ i . (4.2.4a)

After the δ(J 3 ) projection, the only surviving fields are
Φ = a Aa + ic− B , (4.2.4b)

where B is the auxiliary field. Then, using the relations (from (4.1.8))
M 3 a = 0 , M 3 ± = ±± ;
M + a = M + + = 0 , M + − = 2i+ ;
M +a b = η ab + , M +a + =0 , M +a − = −ia ; (4.2.5)

we find the lagrangian and invariance

L = 12 Aa 2 Aa − 2B 2 + 2B∂ a Aa ;

δAa = ∂ a λ , δB = 12 2 λ ; (4.2.6)

which yields the usual result after elimination of B by its equation of motion. This
is the same lagrangian, including signs and auxiliary-field redefinitions, as for the
GL(1)-type 4+4-extended OSp(1,1|2), (4.1.45a).

Any IGL(1) action can be obtained from a corresponding OSp(1,1|2) action, and
vice versa [3.13]. Eliminating ψ from (4.2.3) by its equation of motion,

0= ∼ M + δ(M 3 + 1)ψ − Q+ δ(M 3 )φ
δψ †

→ L = 12 φ† (2 − M 2 − 2Q+ M +−1 Q+ )δ(M 3 )φ , (4.2.7)

the OSp(1,1|2) action (4.1.6) is obtained:

(2 − M 2 − 2Q+ M +−1 Q+ ) δ(M 3 ) − δ(M αβ 2 )

= (2 − M 2 − 2Q+ M +−1 Q+ )M + M +−1 δ(M 3 )

= (2 − M 2 )M + − 2Q+ M +−1 M + Q+ M +−1 δ(M 3 )

= (2 − M 2 )M + − 2Q+2 M +−1 δ(M 3 )


→ L = 12 φ† (2 − M 2 − 2Q+ M +−1 Q+ )δ(M αβ 2 )φ

= 12 φ† (2 − M 2 + Qα 2 )δ(M αβ 2 )φ . (4.2.8)

We have also used Q+2 = (2 − M 2 )M + , which follows from the OSp commutation
relations, or from Q2 = 0. M +−1 is an Sp(2) lowering operator normalized so that
it is the inverse of the raising operator M + , except that it vanishes on states where
M 3 takes its minimum value [4.1]. It isn’t an inverse in the strict sense, since M +
vanishes on certain states, but it’s sufficient for it to satisfy

M + M +−1 M + = M + . (4.2.9)

We can obtain an explicit expression for M +−1 using familiar properties of SO(3)
(SU(2)). The Sp(2) operators are related to the conventionally normalized SO(3)
operators by (M 3 , M ± ) = 2(T 3, T ± ). However, these are really SO(2,1) operators,
and so have unusual hermiticity conditions: T + and T − are each hermitian, while T 3
is antihermitian. Since for any SU(2) algebra T the commutation relations

[T 3 , T ± ] = ±T ± , [T + , T − ] = 2T 3 (4.2.10)


T 2 = (T 3 )2 + 12 {T + , T − } = T (T +1) → T ∓ T ± = (T ∓T 3 )(T ±T 3 +1) , (4.2.11)
4.2. IGL(1) 65

we can write
1 1 − δ T 3 ,T
T +−1 = T −
= T− . (4.2.12)

T T + (T − T )(T + T + 1)
3 3

We can then verify (4.2.9), as well as the identities

T +−1 T + T +−1 = T +−1 ,

T +−1 T + = 1 − δ T 3 ,T , T + T +−1 = 1 − δ T 3 ,−T . (4.2.13)

Conversely, the IGL(1) action can be obtained by partial gauge-fixing of the
OSp(1,1|2) action, by writing L of (4.2.3) as L of (4.2.7) plus a pure BRST variation.
Using the covariantly second-quantized BRST operator of sect. 3.4, we can write

L = L + Q, −φ† M +−1 [Q, δ(M 3 )φ]c . (4.2.14)

Alternatively, we can use functional notation, defining the operator

Q=− dxdc (QΦ) . (4.2.15)
 the extra terms fix the invariance generated by R, which
In terms of J + α = (R, R),
had allowed c to be gauged away. This also breaks the Sp(2) down to GL(1), and
breaks the antiBRST invariance. Another way to understand this is by reformulating
the IGL(1) in terms of a field which has all the zero-modes of the OSp(1,1|2) field Φ.
Consider the action

dD xd2 xα dx− Φ† p+ δ(J 3 )iδ(J −+ )δ(R)δ(Q) Φ . (4.2.16)

The gauge invariance is now given by the 4 generators appearing as arguments of the δ
functions, and is reduced from the OSp(1,1|2) case by the elimination of the generators
 This algebra is the algebra GL(1|1) of N = 2 supersymmetric quantum
(J ± , R, Q).
mechanics (also appearing in the IGL(1) formalism for the closed string [4.10]: see
sect. 8.2): The 2 fermionic generators are the “supersymmetries,” J 3 + J −+ is the
O(2) generator which scales them, and J 3 − J −+ is the “momentum.” If the gauge
coordinates x− and c̃ are integrated out, the action (4.2.1) is obtained, as can be seen
with the aid of (3.4.3).

In contrast to the light cone, where the hamiltonian operator H (= −p− ) is
essentially the action ((2.4.4)), we find that with the new covariant, second-quantized
bracket of (3.4.7) the covariant action is the BRST operator: Because the action

(2.4.8) of a generator (2.4.7) on Φ† is equivalent to the generator’s functional derivative
(because of (3.4.7)), the gauge-invariant action now thought of as an operator satisfies
[S, Φ† ]c = 1
2 . (4.2.17)
Furthermore, since the gauge-covariant equations of motion of the theory are given
by the BRST transformations generated by the operator Q, one has
= −2i[Q, Φ† ]c (4.2.18)
→ S = −iQ . (4.2.19)

(Strictly speaking, S and Q may differ by an irrelevant Φ-independent term.) This
statement can be applied to any formalism with field equations that follow from the
BRST operator, independent of whether it originates from the light-cone, and it holds
in interacting theories as well as free ones. In particular, for the case of interacting
Yang-Mills, the action follows directly from (3.4.18). After restricting the fields to
J 3 = 0, this gives the interacting generalization of the example of (4.2.6). The action
can also be written as S = −2i dΦ QΦ, where dΦ Φn ≡ n+1
Φ† Φn .
This operator formalism is also useful for deriving the gauge invariances of the
interacting theory, in either the IGL(1) or OSp(1,1|2) formalisms (although the cor-
responding interacting action is known in this form only for IGL(1), (4.2.19).) Just
as the global BRST invariances can be written as a unitary transformation (in the
notation of (3.4.17))

U = eiLG , G =
O ,
= constant , (4.2.20)

where O is any IGL(1) (or OSp(1,1|2)) operator (in covariant second-quantized form),
the gauge transformations can be written similarly but with

G = [f, O]c , (4.2.21)

where f is linear in Φ (f = ΛΦ) for the usual gauge transformation (and f ’s higher-
order in Φ may give field-dependent gauge transformations). Thus, Φ = UΦU −1 ,
and g(Φ) = Ug(Φ)U −1 for any functional g of Φ. In the free case, this reproduces
This relation between OSp(1,1|2) and IGL(1) formalisms is important for relating
different first-quantizations of the string, as will be discussed in sect. 8.2.
4.3. Extra modes 67

4.3. Extra modes

As discussed in sect. 3.2, extra sets of unphysical modes can be added to BRST
formalisms, such as those which Lorentz gauges have with respect to temporal gauges,
or those in the 4+4-extended formalisms of sect. 3.6. We now prove the equivalence
of the OSp(1,1|2) actions of formulations with and without such modes [3.13]. Given
that IGL(1) actions and equations of motion can be reduced to OSp forms, it’s suf-
ficient to show the equivalence of the IGL actions with and without extra modes.
The BRST and ghost-number operators with extra modes, after the redefinition of
(3.3.6), differ from those without by the addition of abelian terms. We’ll prove that
the addition of these terms changes the IGL action (4.2.1) only by adding auxiliary
and gauge degrees of freedom. To prove this, we consider adding such terms 2 sets of
modes at a time (an even number of additional ghost modes is required to maintain
the fermionic statistics of the integration measure):

Q = Q0 + (b† f − f † b) ,

[b, g† ] = [g, b† ] = {c, f † } = {f, c† } = 1 , (4.3.1)

in terms of the “old” BRST operator Q0 and the 2 new sets of modes b, g, c,
and f, and their hermitian conjugates. We also assume boundary conditions in the
new coordinates implied by the harmonic-oscillator notation. (Otherwise, additional
unphysical fields appear, and the new action isn’t equivalent to the original one: see
below.) By an explicit expansion of the new field over all the new oscillators,

Φ= (Amn + iB mn c† + iC mn f † + iD mn f † c† ) √ (b† )m (g† )n |0
, (4.3.2)
m,n=0 m!n!

we find

Φ† QΦ = (A† Q0 A + 2B † Q0 C − D † Q0 D) + 2B † (b† D − ibA)

= (A† mn Q0 Anm + 2B † mn Q0 C nm − D † mn Q0 D nm )

√ 1
+ 2B † mn ( nD n−1,m − i √ An,m+1 ) . (4.3.3)

(The ground state in (4.3.2) and the matrix elements evaluated in (4.3.3) are with
respect to only the new oscillators.) We can therefore shift Amn by a Q0 C m,n−1 term
to cancel the B † Q0 C term (using Q0 2 = 0), and then B mn by a Q0 D m,n−1 term to
cancel the D † Q0 D term. We can then shift Amn by D m−1,n−1 to cancel the B † D

term, leaving only the A† Q0 A and B † A terms. (These redefinitions are equivalent to
the gauge choices C = D = 0 using the usual invariance δΦ = QΛ.) Finally, we can
eliminate the Lagrange multipliers B by their equations of motion, which eliminate
all of A except Am0 , and from the form of the remaining A† Q0 A term we find that
all the remaining components of A except A00 drop out (i.e., are pure gauge). This
leaves only the term A† 00 Q0 A00 . Thus, all the components except the ground state
with respect to the new oscillators can be eliminated as auxiliary or gauge degrees of
freedom. The net result is that all the new oscillators are eliminated from the fields
and operators in the action (4.2.1), with Q thus replaced by Q0 (and similarly for J 3 ).
(A similar analysis can be performed directly on the equations of motion QΦ = 0,
giving this general result for the cohomology of Q even in cases when the action is not
given by (4.2.1).) This elimination of new modes required that the creation operators
in (4.3.3) be left-invertible:
1 1
a†−1 a† = 1 → a†−1 = a= a → N ≥0 , (4.3.4)
a† a+1 N +1
implying that all states must be expressible as creation operators acting on a ground
state, as in (4.3.2) (the usual boundary conditions on harmonic oscillator wave func-
tions, except that here b and g correspond to a space of indefinite metric). This
proves the equivalence of the IGL(1) actions, and thus, by the previous argument,
also the OSp(1,1|2) actions, with and without extra modes, and that the extra modes
simply introduce more gauge and auxiliary degrees of freedom.
Such extra modes, although redundant in free theories, may be useful in formulat-
ing larger gauge invariances which simplify the form of interacting theories (as, e.g.,
in nonlinear σ models). The use in string theories of such extra modes corresponding
to the world-sheet metric will be discussed in sect. 8.3.

4.4. Gauge fixing
We now consider gauge fixing of these gauge-invariant actions using the BRST
algebra from the light cone, and relate this method to the standard second-quantized
BRST methods described in sects. 3.1-3 [4.1]. We will find that the first-quantized
BRST transformations of the fields in the usual gauge-fixed action are identical to the
second-quantized BRST transformations, but the first-quantized BRST formalism has
a larger set of fields, some of which drop out of the usual gauge-fixed action. (E.g., see
(3.4.19). However, gauges exist where these fields also appear.) Even in the IGL(1)
formalism, although all the “propagating” fields appear, only a subset of the BRST
4.4. Gauge fixing 69

“auxiliary” fields appear, since the two sets are equal in number in the first-quantized
IGL(1) but the BRST auxiliaries are fewer in the usual second-quantized formalism.
We will also consider gauge fixing to the light-cone gauge, and reobtain the original
light-cone theories to which 2+2 dimensions were added.
For covariant gauge fixing we will work primarily within the IGL(1) formalism,
but similar methods apply to OSp(1,1|2). Since the entire “hamiltonian” 2 − M 2
vanishes under the constraint Q = 0 (acting on the field), the free gauge-fixed action
of the field theory consists of only a “gauge-fixing” term:
1 †
S = i Q, dxdc 2 Φ OΦ

= dxdc 12 Φ† [O, iQ}Φ

= dxdc 12 Φ† KΦ , (4.4.1)

for some operator O, where the first Q, appearing in the covariant bracket, is under-
stood to be the second-quantized one. In order to get 2 − M 2 as the kinetic operator
for part of Φ, we choose
∂ ∂ +
O = − c, → K = c(2 − M 2 ) − 2 M . (4.4.2)
∂c ∂c
When expanding the field in c, 2 −M 2 is the kinetic operator for the piece containing
all physical and ghost fields. Explicitly, (3.4.3b), when substituted into the lagrangian
L = 12 Φ† KΦ and integrated over c, gives

L = 12 φ† (2 − M 2 )φ + ψ † M + ψ , (4.4.3)
and in the BRST transformations δΦ = i
QΦ gives

δφ = i
(Q+ φ − M + ψ) , δψ = i
Q+ ψ − 12 (2 − M 2 )φ) . (4.4.4)

φ contains propagating fields and ψ contains BRST auxiliary fields. Although the
propagating fields are completely gauge-fixed, the BRST auxiliary fields have the
gauge invariance
δψ = λ , M +λ = 0 . (4.4.5)

The simplest case is the scalar Φ = ϕ(x). In this case, all of ψ can be gauged away
by (4.4.5), since M + = 0. The lagrangian is just 12 ϕ(2 − m2 )ϕ. For the massless
vector (cf. (4.2.4b)),
Φ = i Ai + ic− B , (4.4.6)

where we have again used (4.4.5). By comparing (4.4.3) with (4.2.3), we see that the
φ2 term is extended to all J 3 , the ψ 2 term has the opposite sign, and the crossterm
is dropped. We thus find

L = 12 (Ai )† 2 Ai + 2B 2 = 12 Aa 2 Aa − iC 2 C + 2B 2 , (4.4.7)
 due to (3.4.4,13). This agrees with the result
where we have written Aα = i(C, C),
 The BRST transformations (4.4.4),
(3.2.11) in the gauge ζ = 1, where this B = 12 B.
using (4.2.5), are

δAa = i
∂ a C ,

δC = 0 ,

δ C =
(2B − ∂ · A) ,

δB = i 12
2 C , (4.4.8)

which agrees with the linearized case of (3.2.8).
We next prove the equivalence of this form of gauge fixing with the usual ap-
proach, described in sect. 3.2 [4.1] (as we have just proven for the case of the massless
vector). The steps are: (1) Add terms to the original BRST auxiliary fields, which
vanish on shell, to make them BRST invariant, as they are in the usual BRST for-
 in (3.2.11).)
mulation of field theory. (In Yang-Mills, this is the redefinition B → B
(2) Use the BRST transformations to identify the physical fields (which may include
auxiliary components). We can then reobtain the gauge-invariant action by dropping
all other fields from the lagrangian, with the gauge transformations given by replacing
the ghosts in the BRST transformations by gauge parameters.
In the lagrangian (4.4.3) only the part of the BRST auxiliary field ψ which appears
in M + ψ occurs in the action; the rest of ψ is pure gauge and drops out of the action.
Thus, we only require that the shifted M + ψ be BRST-invariant:

ψ = ψ + Aφ , δM + ψ = 0 . (4.4.9)

Using the BRST transformations (4.4.4) and the identities (from (3.4.3b))

Q2 = 0 → [2 − M 2 , M + ] = [2 − M 2 , Q+ ] = [M + , Q+ ] = 0 ,

Q+2 = 12 (2 − M 2 )M + , (4.4.10)

we obtain the conditions on A:

(Q+ − M + A)Q+ = (Q+ − M + A)M + = 0 . (4.4.11)
4.4. Gauge fixing 71

The solution to these equations is

A = M +−1 Q+ . (4.4.12)

Performing the shift (4.4.9), the gauge-fixed lagrangian takes the form

L = 12 φ† K̂φ + 2ψ† (1 − δ M 3 ,−2T )Q+ φ + ψ† M + ψ , (4.4.13a)
K̂ = 2 − M 2 − 2Q+ M +−1 Q+ , (4.4.13b)

where T is the “isospin,” as in (4.2.11), for M αβ . The BRST transformations can
now be written as

δδ M 3 ,−2T φ =
δ M 3 ,−2T Q+ φ , δ(1 − δ M 3 ,−2T )φ = −
M + ψ ,

δ ψ =
δ M 3 ,2T (Q+ ψ − 12 K̂φ) . (4.4.14)

The BRST transformation of ψ is pure gauge, and can be dropped. (In some of
the manipulations we have used the fact that Q, Q+ , and 2 − M 2 are symmetric,
i.e., even under integration by parts, while M + is antisymmetric, and Q and Q are
antihermitian while M + and 2 − M 2 are hermitian. In a coordinate representa-
tion, particularly for c, all symmetry generators, such as Q, Q+ , and M + , would be
antisymmetric, since the fields would be real.)
 but we
We can now throw away the BRST-invariant BRST auxiliary fields ψ,
must also separate the ghost fields in φ from the physical ones. According to the
usual BRST procedure, the physical modes of a theory are those which are both
BRST-invariant and have vanishing ghost number (as well as satisfy the field equa-
tions). In particular, physical fields may transform into ghosts (corresponding to
gauge transformations, since the gauge pieces are unphysical), but never transform
into BRST auxiliary fields. Therefore, from (4.4.14) we must require that the phys-
ical fields have M 3 = −2T to avoid transforming into BRST auxiliary fields, but we
also require vanishing ghost number M 3 = 0. Hence, the physical fields (located in
φ) are selected by requiring the simple condition of vanishing isospin T = 0. If we
project out the ghosts with the projection operator δ T 0 and use the identity (4.2.8),
we obtain a lagrangian containing only physical fields:

L1 = 12 φ† (2 − M 2 + Q2 )δ T 0 φ . (4.4.15)

Its gauge invariance is obtained from the BRST transformations by replacing the
ghosts (the part of φ appearing on the right-hand side of the transformation law)

with the gauge parameter (the reverse of the usual BRST quantization procedure),
and we add gauge transformations to gauge away the part of φ with T = 0:

δφ = −i 12 Qα Λα + 12 M αβ Λαβ , (4.4.16)

where we have obtained the Q− term from closure of Q+ with M αβ . The invariance
of (4.4.15) under (4.4.16) can be verified using the above identities. This action and
invariance are just the original ones of the OSp(1,1|2) gauge-invariant formalism (or
the IGL(1) one, after eliminating the NL auxiliary fields). The gauge-fixing functions
for the Λ transformations are also given by the BRST transformations: They are the
transformations of ghosts into physical fields:

F GF = Q+ δ T 0 φ = pa (M + a δ T 0 φ) + M(M + m δ T 0 φ) . (4.4.17)

(The first term is the usual Lorentz-gauge gauge-fixing function for massless fields,
the second term the usual addition for Stueckelberg/Higgs fields.) The gauge-fixed
lagrangian (physical fields only) is thus

LGF = L1 − 14 F GF † M − F GF = 12 φ† (2 − M 2 )δ T 0 φ , (4.4.18)

in agreement with (4.4.3).

In summary, we see that this first-quantized gauge-fixing procedure is identical
to the second-quantized one with regard to (1) the physical gauge fields, their gauge
transformations, and the gauge-invariant action, (2) the BRST transformations of the
physical fields, (3) the closure of the BRST algebra, (4) the BRST invariance of the
gauge-fixed action, and (5) the invertibility of the kinetic operator after elimination
of the NL fields. (1) implies that the two theories are physically the same, (2) and
(3) imply that the BRST operators are the same, up to additional modes as in sect.
4.3, (4) implies that both are correctly gauge fixed (but perhaps in different gauges),
and (5) implies that all gauge invariances have been fixed, including those of ghosts.
Concerning the extra modes, from the 2 −M 2 form of the gauge-fixed kinetic operator
we see that they are exactly the ones necessary to give good high-energy behavior of
the propagator, and that we have chosen a generalized Fermi-Feynman gauge. Also,
note the fact that the c = 0 (or xα = 0) part of the field contains exactly the right
set of ghost fields, as was manifest by the arguments of sect. 2.6, whereas in the usual
second-quantized formalism one begins with just the physical fields manifest, and the
ghosts and their ghosts, etc., must be found by a step-by-step procedure. Thus we
see that the OSp from the light cone not only gives a straightforward way for deriving
4.4. Gauge fixing 73

general free gauge-invariant actions, but also gives a method for gauge fixing which
is equivalent to, but more direct than, the usual methods.

We now consider gauge fixing to the light cone. In this gauge the gauge theory
reduces back to the original light-cone theory from which it was heuristically obtained
by adding 2+2 dimensions in sects. 3.4, 4.1. This proves a general “no-ghost theo-
rem,” that the OSp(1,1|2) (and IGL(1)) gauge theory is equivalent on-shell to the
corresponding light-cone theory, for any Poincaré representation (including strings as
a special case).

Consider an arbitrary bosonic gauge field theory, with action (4.1.6). (Fermions
will be considered in the following section.) Without loss of generality, we can choose
M 2 = 0, since the massive action can be obtained by dimensional reduction. The
light-cone gauge is then described by the gauge-fixed field equations

p2 φ = 0 (4.4.19a)

subject to the gauge conditions, in the Lorentz frame pa = δ a + p+ ,

M αβ φ = 0 , Qα φ = M α− p+ φ = 0 , (4.4.19b)

with the residual part of the gauge invariance

δφ = −i 12 Qα Λα ∼ M α− Λα , (4.4.19c)

where ± now refer to the usual “longitudinal” Lorentz indices. (The light-cone gauge
is thus a further gauge-fixing of the Landau gauge, which uses only (4.4.19ab).)
(4.4.19bc) imply that the only surviving fields are singlets of the new OSp(1,1|2)
algebra generated by M αβ , M α± , M −+ (with longitudinal Lorentz indices ±): i.e.,
those which satisfy M AB φ = 0 and can’t be gauged away by δφ = M BA ΛAB .

We therefore need to consider the subgroup SO(D−2)⊗OSp(1,1|2) of OSp(D−1,1|2)
(the spin group obtained by adding 2+2 dimensions to the original SO(D−2)), and
determine which parts of an irreducible OSp(D−1,12) representation are OSp(1,1|2)
singlets. This is done most simply by considering the corresponding Young tableaux
(which is also the most convenient method for adding 2+2 dimensions to the original
representation of the light-cone SO(D−2)). This means considering tensor prod-
ucts of the vector (defining) representation with various graded symmetrizations and
antisymmetrizations, and (graded) tracelessness conditions on all the indices. The
obvious OSp(1,1|2) singlet is given by allowing all the vector indices to take only

SO(D−2) values. However, the resulting SO(D−2) representation is reducible, since
it is not SO(D−2)-traceless. The OSp(D−1,1|2)-tracelessness condition equates its
SO(D−2)-traces to OSp(1,1|2)-traces of representations which differ only by replacing
the traced SO(D−2) indices with traced OSp(1,1|2) ones. However, OSp(1,1|2) (or
OSp(2N|2N) more generally) has the unusual property that its traces are not true
singlets. The simplest example [4.11] (and one which we’ll show is sufficient to treat
the general case) is the graviton of (4.1.15). Considering just the OSp(1,1|2) values
of the indices, there are 2 states which are singlets under the bosonic subgroup GL(2)
generated by M αβ , M −+ , namely (+ −) , |α

. However, of these two states, one
linear combination is pure gauge and one is pure auxiliary:

 φ1 = A A → M AB φ1 = 0 , but
 φ1= − 12 M ± α (∓ α) ,
 φ2 = (+ −) − α α →  φ2 = M AB ΛBA , but
M ±α φ2 = −2(± α) = 0 . (4.4.20)

This result is due basically to the fact that the graded trace can’t be separated out
in the usual way with the metric because of the identity

η AB η BA = η A A = (−1)A δ A A = 2 − 2 = 0 . (4.4.21)

Similarly, the reducible OSp(1,1|2) representation which consists of the unsymmetrized
direct product of an arbitrary number of vector representations will contain no sin-
glets, since any one trace reduces to the case just considered, and thus the represen-
tations which result from graded symmetrizations and antisymmetrizations will also
contain none. Thus, no SO(D−2)-traces of the original OSp(D−1,1|2) representation
need be considered, since they are equated to OSp(1,1|2)-nonsinglet traces by the
OSp(D−1,1|2)-tracelessness condition. Hence, the only surviving SO(D−2) repre-
sentation is the original irreducible light-cone one, obtained by restricting all vector
indices to their SO(D−2) values and imposing SO(D−2)-tracelessness.

These methods apply directly to open strings. The modification for closed strings
will be discussed in sect. 11.1.
4.5. Fermions 75

4.5. Fermions

The results of this section are based on the OSp(1,1|2) algebra of sect. 3.5.

The action and invariances are again given by (4.1.1), with the modified J AB of
 . We also allow for the inclu-
sect. 3.5, and (4.1.3) is unchanged except for M αβ → M αβ

sion of a matrix factor in the Hilbert-space metric to maintain the (pseudo)hermiticity
of the spin operators (e.g., γ 0 for a Dirac spinor, so Φ† → Φ̄ = Φ† γ 0 ). Under the
action of the δ functions, we can make the replacement

p+ 2 J −α 2 → − 34 (p2 + M 2 ) + (M α b pb + M αm M)2

− 12 γ̃ α (M α b pb + M αm M) γ̃ − + γ̃ + 21 (p2 + M 2 ) , (4.5.1)

where p2 ≡ pa pa . Under the action of the same δ functions, the gauge transformation
generated by J −α is replaced with
δΦ = J − α Λα → (M αb pb + M α m M) − 12 γ̃ α γ̃ − + γ̃ + 21 (p2 + M 2 ) Λα .
Choosing Λα = γ̃ α Λ, the γ̃ − part of this gauge transformation can be used to choose
the partial gauge
γ̃ + Φ = 0 . (4.5.3)

The action then becomes

S=  2 )iγ̃ α (M b p + M M) φ ,
dD x φ̄ δ(M (4.5.4)
αβ α b αm

where we have reduced Φ to the half represented by φ by using (4.5.3). (The γ̃ ± can
be represented as 2×2 matrices.) The remaining part of (4.5.2), together with the
J αβ transformation, can be written in terms of γ̃ ± -independent parameters as

δφ = (M αb pb + M α m M)(Aα + 12 γ̃ α γ̃ β Aβ ) + − 14 (p2 + M 2 ) + (M α b pb + M αm M)2 B
 αβ Λ
+ 12 M . (4.5.5)

One way to get general irreducible spinor representations of orthogonal groups
(except for chirality conditions) is to take the direct product of a Dirac spinor with an
irreducible tensor representation, and then constrain it by setting to zero the result of
multiplying by a γ matrix and contracting vector indices. Since the OSp representa-
tions used here are obtained by dimensional continuation, this means we use the same
constraints, with the vector indices i running over all commuting and anticommuting

values (including m, if we choose to define M im by dimensional reduction from one
extra dimension). The OSp spin operators can then be written as

M ij = M̌ ij + 14 [γ i , γ j } , (4.5.6)

where M̌ are the spin operators for some tensor representation and γ i are the OSp
Dirac matrices, satisfying the OSp Clifford algebra

{γ i , γ j ] = 2η ij . (4.5.7)

We choose similar relations between γ’s and γ̃’s:

{γ i , γ̃ B ] = 0 . (4.5.8)

Then, noting that 12 (γ α +iγ̃ α ) and 12 (γ α −iγ̃ α ) satisfy the same commutation relations
as creation and annihilation operators, respectively, we define

γ α = aα + a† α , γ̃ α = i aα − a† α ; aα , a†β = δ α β . (4.5.9)

We also find
 = M̌
M †
αβ αβ + a (α aβ) . (4.5.10)

This means that an arbitrary representation ψ (α···β) of the part of the Sp(2) generated

by M̌ αβ that is also a singlet under the full Sp(2) generated by M αβ can be written

ψ (α···β) = a† α · · · a† β ψ , aα ψ = 0 . (4.5.11)

In particular, for a Dirac spinor M̌ = 0, so the action (4.5.4) becomes simply (see
also (4.1.40)) 
S= dD x φ̄(/
p + M)φ
/ , (4.5.12)

/ ≡ γ m M (γ m is like γ 5 ), all dependence on γ α and γ̃ α has
where p/ ≡ γ a pa , M
been eliminated, and the gauge transformation (4.5.5) vanishes. (The transformation
/ → p/ + iM.)
φ → eiγ m π/4 φ takes p/ + M

In the case of the gravitino, we start with φ = |i
φi , where |i
is a basis for
 φ = 0 but also the
the representation of M̌ (only). φ must satisfy not only M αβ

irreducibility condition

γ i φi = 0 → φα = 12 γ α γ a φa , aα φa = 0 . (4.5.13)
4.5. Fermions 77

(a† α or iγ̃ α could be used in place of γ α in the solution for φα .) Then using (4.1.8)
for M̌ on φ, straightforward algebra gives the action

S = dD x φ̄a (η ab p/ − p(a γ b) + γ a p/γ b )φb

= dD x φ̄a γ abc pb φc , (4.5.14)

where γ abc = (1/3!)γ [a γ b γ c] , giving the usual gravitino action for arbitrary D. The
gauge invariance remaining in (4.5.5) after using (4.5.13-14), and making suitable
redefinitions, reduces to the usual

δφa = ∂ a λ . (4.5.15)

We now derive an alternative form of the fermionic action which corresponds to
actions given in the literature for fermionic strings. Instead of explicitly solving the

constraint M 
αβ = 0 as in (4.5.11), we use the M αβ gauge invariance of (4.5.5) to

“rotate” the aα† ’s. For example, writing aα = (a+ , a− ), we can rotate them so they
all point in the “+” direction: Then we need consider only φ’s of the form φ(a+† ) |0
 2 ) then
(The + value of α should not be confused with the + index on p+ .) The δ(M αβ

picks out the piece of the form (4.5.11). (It “smears” over directions in Sp(2). This
use of aα is similar to the “harmonic coordinates” of harmonic superspace [4.12].) We
can also pick
+† a−†
φ = e−ia φ(a+† ) |0
, (4.5.16a)
 projection) just redefines some components by
since the exponential (after δ(M)
shifting by components of lower M̌ -spin. In this gauge, writing γ α = (s, u), γ̃ α =
(s̃, ũ), we can rewrite φ as
 +† a−†  
0̃ = e−ia |0
↔ s̃ 0̃ = u 0̃ = 0 ,

φ = φ(s) 0̃ ↔ s̃φ = 0 . (4.5.16b)

By using an appropriate (indefinite-metric) coherent-state formalism, we can choose
s to be a coordinate (and u = −2i∂/∂s). We next make the replacement
 2 + ∂ 2 3
δ(M αβ ) → ds µ(s) δ(2M̌ + s )δ M̌ + s (4.5.17)

after pushing it to the right in (4.5.4) so it hits the φ, where we have just replaced
 2 = 0 with M
projection onto M + = M
 3 = 0 (which implies M
 − = 0). The first δ

function factor is a Dirac δ function, while the second is a Kronecker δ. µ(s) is an

appropriate measure factor; instead of determining it by an explicit use of coherent
states, we fix it by comparison with the simplest case of the Dirac spinor: Using

(s)δ(s2 + r 2 )(a + bs) = b , (4.5.18a)

we find
µ(s) ∼
(s) . (4.5.18b)
Then only the ũ term of (4.5.4) contributes in this gauge, and we obtain (the ũ itself
having already been absorbed into the measure (4.5.18b))
+ + 2 ∂ 3
S = −2 d x ds
(s) φ̄ Q δ(2M̌ + s )δ M̌ + s
φ . (4.5.19)
(All dependence on γ α and γ̃ α has been reduced to φ being a function of just s. This
action was first proposed for the string [4.13].) Since the first δ function can be used
to replace any s2 with a −2M̌ + , we can perform all such replacements, or equivalently
choose the gauge
φ(s) = φ0 + sφ1 . (4.5.20)
(An equivalent procedure was performed for the string in [4.14].) For the Dirac spinor,
after integration over s (including that in Q+ = 12 s(/
p + M)),
/ the Dirac action is easily
found (φ1 drops out). For general spinor representations, Q+ has an additional Q̌+
term, and s integration gives the lagrangian

L = φ̄0 (/
p + M)δ(
/ M̌ 3 )φ0 + 2 φ̄1 Q̌+ δ(M̌ 3 )φ0 − φ̄0 Q̌+ δ(M̌ 3 + 1)φ1 . (4.5.21)

However, γ-matrix trace constraints (such as (4.5.13)) must still be solved to relate
the components.
The explicit form of the OSp(1,1|2) operators for the fermionic string to use with
these results follows from the light-cone Poincaré generators which will be derived in
sect. 7.2. The s dependence of Q+ is then slightly more complicated (it also has a
∂/∂s term). (The resulting action first appeared in [4.14,15].)
The proof of equivalence to the light cone is similar to that for bosons in the
previous section. Again considering the massless case, the basic difference is that we
now have to use, from (3.5.3b),

M  =M p +S
αβ αβ + S αβ , α α− + −α , (4.5.22)

and other corresponding generators, as generating the new OSp(1,1|2). This is just
the diagonal OSp(1,1|2) obtained from S AB and the one used in the bosonic case.
Exercises 79

In analogy to the bosonic case, we consider reducible OSp(D−1,1|2) representations
corresponding to direct products of arbitrary numbers of vector representations with
one spinor representation (represented by graded γ-matrices). We then take the direct
product of this with the S AB representation, which is an OSp(1,1|2) spinor but an
SO(D−2) scalar. Since the direct product of 2 OSp(1,1|2) spinors gives the direct
sum of all (graded) antisymmetric tensor representations, each once (by the usual
γ-matrix decomposition), from the bosonic result we see that the only way to get an
OSp(1,1|2) singlet is if all vector indices again take only their SO(D−2) values. The
OSp(D−1,1|2) spinor is the direct product of an SO(D−2) spinor with an OSp(1,1|2)
spinor, so the net result is the original light-cone one. In the bosonic case traces in
OSp(1,1|2) vector indices did not give singlets because of (4.4.21); a similar result
holds for γ-matrix traces because of

γ Aγ A = ηAA = 0 . (4.5.23)

More general representations for S AB could be considered, e.g., as in sect. 3.6. The
 and Q
action can then be rewritten as (4.1.6), but with M αβ and Qα replaced by M 
αβ α

of (3.5.3b). In analogy to (4.5.2,3), the S −α part of the J −α transformation can be used
to choose the gauge S +α Φ = 0. Then, depending on whether the representation allows
application of the “lowering” operators S −α 0,1, or 2 times, only the terms of zeroth,
first, or second order in S −α , respectively, can contribute in the kinetic operator.
Since these terms are respectively second, first, and zeroth order in derivatives, they
can be used to describe bosons, fermions, and auxiliary fields.
The argument for equivalence to the light cone directly generalizes to the U(1)-

type 4+4-extended OSp(1,1|2) of sect. 3.6. Then M AB = M AB + S AB has a singlet

only when M AB and S AB are both singlets (for bosons) or both Dirac spinors (for

(1) Derive (4.1.5).
(2) Derive (4.1.6).
(3) Find the gauge-invariant theory resulting from the light-cone theory of a totally
symmetric, traceless tensor of arbitrary rank.
(4) Find the explicit infinitesimal gauge transformations of ea m , em a , e−1 , g mn , g mn ,
and ω abc from (4.1.20-22). Linearize, and show the gauge e[am] = 0 can be ob-

tained with λab . Find the transformation for a covariant vector Aa (from a simi-
larity transformation, like (4.1.22)).

(5) Write cabc explicitly in terms of ea m . Find T abc and Rabcd in terms of cabc and
ω abc . Derive (4.1.31). Linearize to get (4.1.17).

(6) Find an expression for ω abc when (4.1.24) is not imposed, in terms of T abc and
the ω of (4.1.27).

(7) Derive global Poincaré transformations by finding the subgroup of (4.1.20) which
leaves (4.1.28) invariant.

(8) Find the field equation for φ from (4.1.36), and show that (4.1.38) satisfies it.

(9) Derive the gauge-covariant action for gravity in the GL(1)-type 4+4-extension of
OSp(1,1|2), and compare with the U(1) result, (4.1.47).

(10) Find the BRST transformations for the IGL(1) formalism of sect. 4.2 (BRST1,
derived from the light cone) for free gravity. Find those for the usual IGL(1) for-
malism of sect. 3.2 (BRST2, derived from second-quantizing the gauge-invariant
field theory). After suitable redefinitions of the BRST1 fields (including aux-
iliaries and ghosts), show that a subset of these fields that corresponds to the
complete set of fields in the BRST2 formalism has identical BRST transforma-

(11) Formulate φ3 theory as in (4.2.19), using the bracket of (3.4.7).

(12) Derive the gauge transformations for interacting Yang-Mills by the covariant
second-quantized operator method of (4.2.21), in both the IGL(1) and OSp(1,1|2)

(13) Find the free gauge-invariant action for gravity in the IGL(1) formalism, and
compare with the OSp(1,1|2) result (4.1.17). Find the gauge-fixed action by

(14) Perform IGL(1) gauge-fixing, as in sect. 4.4, for a second-rank antisymmetric
tensor gauge field. Perform the analogous gauge fixing by the method of sect.
3.2, and compare. Note that there are scalar commuting ghosts which can be
interpreted as the ghosts for the gauge invariances of the vector ghosts (“ghosts
for ghosts”).

(15) Derive (4.5.14,15).
5.1. Bosonic 81


5.1. Bosonic

If coordinates are considered as fields, and their arguments as the coordinates of
small spacetimes, then the mechanics of particles and strings can be considered as
1- and 2-dimensional field theories, respectively (see sect. 1.1). (However, to avoid
confusion, we will avoid referring to mechanics theories as field theories.) Thus, the
particle is a useful analog of the string in one lower “dimension”, and we here review
its properties that will be found useful below for the string.
As described in sect. 3.1, the mechanics action for any relativistic particle is
completely determined by the constraints it satisfies, which are equivalent to the free
equations of motion of the corresponding field theory. The first-order (hamiltonian)
form ((3.1.10)) is more convenient than the second-order one because (1) it makes
canonical conjugates explicit, (2) the inverse propagator (and, in more general cases,
all other operator equations of motion) can be directly read off as the hamiltonian, (3)
path-integral quantization is easier, and (4) treatment of the supersymmetric case is
clearer. The simplest example is a massless, spinless particle, whose only constraint
is the Klein-Gordon equation p2 = 0. From (3.1.10), the action [5.1] can thus be
written in first-order form as


dτ L ,
L = x · p − g 21 p2 , (5.1.1)

. = ∂/∂τ . The action is invariant under Poincaré trans-
where τ is the proper time, of which the position x, momentum p, and 1-dimensional
metric g are functions, and
formations in the higher-dimensional spacetime described by x, as well as 1D general
coordinate transformations (τ -reparametrizations). The latter can be obtained from
δx = ζp , δp = 0 , δg = ζ
. . (5.1.2)

These differ from the usual transformations by terms which vanish on shell: In general,
any action with more than one field is invariant under δφi = λij δS/δφj , where λij is

antisymmetric. Such invariances may be necessary for off-shell closure of the above
algebra, but are irrelevant for obtaining the field theory from the classical mechanics.
(In fact, in the component formalism for supergravity, gauge invariance is more easily
proven using a first-order formalism with the type of transformations in (5.1.2) rather
than the usual transformations which follow from the second-order formalism [5.2].)
In this case, if we add the transformations
δS δS
δx =
, δ  p = −
δp δx
to (5.1.2), and choose
= g −1ζ, we obtain the usual general coordinate transforma-
tions (see sect. 4.1)
. . .
δ  x =
x , δ  p =
p , δ  g = (
g) . (5.1.4)

The second-order form is obtained by eliminating p:

L = g −1 12 x2
. . (5.1.5)

The transformations (5.1.4) for x and g also follow directly from (5.1.2) upon elimi-
nating p by its equation of motion. The massive case is obtained by replacing p2 with
p2 + m2 in (5.1.1). When the additional term is carried over to (5.1.5), we get
L = 12 g −1 x2 − 12 gm2 . (5.1.6)

g can now also be eliminated by its equation of motion, producing

S = −m dτ −x2 = −m
−dx2 , (5.1.7)

which is the length of the world line.
Besides the 1D invariance of (5.1.1) under reparametrization of τ , it also has the
discrete invariance of τ reversal. If we choose x(τ ) → x(−τ ) under this reversal,
then p(τ ) → −p(−τ ), and thus this proper-time reversal can be identified as the
classical mechanical analog of charge (or complex) conjugation in field theory [5.3],
where φ(x) → φ*(x) implies φ(p) → φ*(−p) for the fourier transform. (Also, the
electromagnetic coupling q dτ x · A(x) when added to (5.1.1) requires the charge
q → −q.)
There are two standard gauges for quantizing (5.1.1). In the light-cone formalism
the gauge is completely fixed (for p+ = 0, up to global transformations, which are
eliminated by boundary conditions) by

x+ = p+ τ . (5.1.8)
5.1. Bosonic 83

We then eliminate p− as a lagrange multiplier with field equation g = 1. The la-
grangian then simplifies to
. .
L = x− p+ + xi pi − 12 (pi 2 + m2 ) , (5.1.9)

with (retarded) propagator
1 2 +m2 )
−iΘ(τ )eiτ 2 (pi (5.1.10a)

(where Θ(τ ) = 1 for τ > 0 and 0 otherwise) or, fourier transforming with respect to
1 1 1
∂ 1 2 2
= 1 2 2
= 1 2 2
. (5.1.10b)
i ∂τ + 2 (pi + m ) + i
p+ p− + 2 (pi + m ) + i
2 (p + m ) + i

For interacting particles, it’s preferable to choose

x+ = τ , (5.1.11)

so that the same τ coordinate can be used for all particles. Then g = 1/p+ , so
the hamiltonian 12 (pi 2 + m2 ) gets replaced with (pi 2 + m2 )/2p+ , which more closely
resembles the nonrelativistic case. If we also use the remaining (global) invariance of
τ reparametrizations (generated by p2 ), we can choose the gauge x+ = 0, which is
the same as choosing the Schrödinger picture.

Alternatively, in the covariant formalism one chooses the gauge

g = constant , (5.1.12)

where g can’t be completely gauge-fixed because of the invariance of the 1D volume

T = dτ g. The functional integral over g is thus replaced by an ordinary integral
over T [5.4], and the propagator is [5.3,5]
 ∞ 1 2 +m2 ) 1
−i dT Θ(T )eiT 2 (p = 1 2 . (5.1.13)
0 2 (p + m2 ) + i

The use of the mechanics approach to the particle is somewhat pointless for the
free theory, since it contains no information except the constraints (from which it
was derived), and it requires treatment of the irrelevant “off-shell” behavior in the
“coordinate” τ . However, the proper-time is useful in interacting theories for studying
certain classical limits and various properties of perturbation theory. In particular,
the form of the propagator given in (5.1.13) (with Wick-rotated τ : see sect. 2.5) is
the most convenient for doing loop integrals using dimensional regularization: The

momentum integrations become simple Gaussian integrals, which can be trivially
evaluated in arbitrary dimensions by analytic continuation from integer ones:
  D/2  D/2
D −ap2 2 −ap2 π
d pe = d pe = . (5.1.14)
(The former integral factors into 1-dimensional ones; the latter is easily performed in
polar coordinates.) The Schwinger parameters τ are then converted into the usual
∞ !
Feynman parameters α by inserting unity as 0 dλ δ(λ − τ ), rescaling τ i = λαi ,
and integrating out λ, which now appears in standard Γ-function integrals, to get
the usual Feynman denominators. An identical procedure is applied in string theory,
but writing the parameters as x = e−τ , w = e−λ . (See (9.1.10).) By not converting
the τ ’s into α’s, the high-energy behavior of scattering amplitudes can be analyzed
more easily [5.6]. Also, the singularities in an amplitude correspond to classical paths
of the particles, and this identification can be seen to be simply the identification
of the τ parameters with the classical proper-time [5.7]. 1-loop calculations can be
performed by introducing external fields (see also sect. 9.1) and treating the path
of the particle in spacetime as closed [5.5,8]. Such calculations can treat arbitrary
numbers of external lines (or nonperturbative external fields) for certain external
field configurations (such as constant gauge-field strengths). Finally, the introduction
of such expressions for propagators in external fields allows the study of classical
limits of quantum field theories in which some quantum fields (represented by the
external field) become classical fields, as in the usual classical limit, while other fields
(represented by the particles described by the mechanics action) become classical
particles [5.9].
This classical mechanics analysis will be applied to the string in chapt. 6.

5.2. BRST
In this section we’ll apply the methods of sect. 3.2-3 to study BRST quantization
of particle mechanics, and find results equivalent to those obtained by more general
methods in sect. 3.4.
In the case of particle mechanics (according to sect. 3.1), for the action of the
previous section we have G = −i 12 (p2 + m2 ), and thus [4.4], for the “temporal” gauge
g = 1, from (3.2.6)
Q = −ic 12 (p2 + m2 ) , (5.2.1)
which agrees with the general result (3.4.3b). We could write c = ∂/∂C so that
in the classical field theory which follows from the quantum mechanics the field
5.2. BRST 85

φ(x, C) could be real (see sect. 3.4). This also follows from the fact that the (τ -
reparametrization) gauge-parameter corresponding to c carries a (proper-)time index
(it’s a 1D vector), and thus changes sign under τ -reversal (mechanics’ equivalent of
field theory’s complex conjugation), and so c is a momentum (φ(x, C) = φ*(x, C),
φ(p, c) = φ*(−p, −c)).
We now consider extending IGL(1) to OSp(1,1|2) [3.7]. By (3.3.2),

Qα = −ixα 21 (p2 + m2 ) − b∂ α . (5.2.2)

In order to compare with sect. 3.4, we make the redefinitions (see (3.6.8))

b=i , g = 12 p+ 2 , (5.2.3a)
(where g is the world-line metric) and the unitary transformation

ln U = −(ln p+ ) 12 [xα , ∂ α ] , (5.2.3b)

1 2

UQα U −1 = −ixα p + m2 + pα pα − ix− pα , (5.2.4)
which agrees with the expression given in (3.4.2) for the generators J −α for the case
of the spinless particle, as does the rest of the OSp(1,1|2) obtained from (3.3.7).
In a lagrangian formalism, for the action (5.1.6) with invariance (5.1.4), (3.3.2)
gives the BRST transformation laws

Qα xa = xα xa
. ,
Qα g = (xα g) ,
Qα xβ = 12 x(α xβ) − C αβ b ,

Qα b = 1
. . .
(x b − bx ) − (x
α α 1 2 α
x + xβ 2 x α ) .
.. (5.2.5)
2 4 β

We first make the redefinition
b̃ = b − 12 (xα 2 ) (5.2.6)
to simplify the transformation law of xα and thus b:
Qα xβ = xα xβ − C αβ b̃ ,

Qα b̃ = xα b . (5.2.7)

We then make further redefinitions
xα → g −1 xα , b̃ → g −1 b − 2(g −1 xα 2 ) , (5.2.8)

which simplify the g transformation, allowing a further simplification for b:
Qα xa = xα g −1 xa = xα pa ,

Qα g = xα
. ,

Qα xβ = −C αβ b ,

Qα b = 0 . (5.2.9)

To get just a BRST operator (as for the IGL(1) formalism), we can restrict the Sp(2)
indices in (5.2.9) to just one value. Then xα for the other value of α (the antighost)
and b can be dropped. (They form an independent IGL(1) multiplet, as described in
sect. 3.2.)

To get the OSp(1,1|2) formalism, we choose a “Lorentz” gauge. We then quantize
with the ISp(2)-invariant gauge-fixing term
. .
L1 = −Qα 2 f (g) = f  (g)(gb − xα 2 ) (5.2.10)

for some arbitrary function f such that f  = 0. Canonically quantizing (where f  (g)
is conjugate to b), and using the equations of motion, we find Qα from its Noether
current (which in D = 1 is also the charge) to be given by (5.2.2). For an IGL(1)
formalism, we can use the temporal gauge (writing xα = (c, c̃))

L1 = iQ[c̃f (g)] = bf (g) − if  (g)c̃c .
. (5.2.11)

(Compare the discussions of gauge choices in sect. 3.2-3.)

Although Lorentz-gauge quantization gave a result equivalent to that obtained
from the light cone in sect. 3.4, we’ll find in sect. 8.3 for the string a result equivalent
to that obtained from the light cone in sect. 3.6.

5.3. Spinning

The mechanics of a relativistic spin-1/2 particle [5.1] is obtained by symmetrizing
the particle action for a spinless particle with respect to one-dimensional (local)
supersymmetry. We thus generalize x(τ ) → X(τ, θ), etc., where θ is a single, real,
anticommuting coordinate. We first define global supersymmetry by the generators
∂ ∂
q= − θi∂ , i∂ ≡ i = −q 2 , (5.3.1a)
∂θ ∂τ
5.3. Spinning 87

which leave invariant the derivatives

d= + θi∂ , ∂ = −id2 . (5.3.1b)
The local invariances are then generated by (expanding covariantly)

K = κid + ki∂ , (5.3.2a)

which act covariantly (i.e., as ( ) = eiK ( )e−iK ) on the derivatives

D = Gd + G∂ , −iD 2 . (5.3.2b)

This gives the infinitesimal transformation

δD = i[K, D] = i(KG − iDκ)d + i(KG − iDk)∂ + 2iκG∂ . (5.3.2c)

We next use κ by the last part of this transformation to choose the gauge

G = 0 → κ = i 12 dk . (5.3.3)

The action (5.1.1) becomes

S = dτ dθ G−1 (−iD 2 X) · P − 12 P · DP

= dτ dθ −iG(dX) · (dP ) − 12 P · dP . (5.3.4)

When expanded in components by dθ → d, and defining

X=x , DX = iγ ;

P =ζ , DP = p ;

G = g −1/2 , dG = ig −1 ψ ; (5.3.5)

when evaluated at θ = 0 (in analogy to sect. 3.2), we find


. .
x · p + iψγ · p − iγ · ζ − g 21 p2 + 12 iζ · ζ
. . (5.3.6)

The (g, x, p) sector works as for the bosonic case. In the (ψ, γ, ζ) sector we see that
the quantity i(γ − 12 ζ) is canonically conjugate to ζ, and thus

γ= + 1ζ , (5.3.7a)
∂ζ 2

which has γ-matrix type commutation relations. It anticommutes with

γ̂ = γ − ζ = − 12 ζ , (5.3.7b)
which is an independent set of γ-matrices. However, it is γ which appears in the
Dirac equation, obtained by varying ψ.
In a light-cone formalism, we again eliminate all auxiliary “−” components by
their equations of motion, and use the gauge invariance (5.3.2-3) to fix the “+”
X + = p+ τ → x+ = p+ τ , γ+ = 0 . (5.3.8)

We then find G = 1, and (5.3.6) reduces to
. . . .
L = x− p+ + xi pi − 12 pi 2 − i(γ − − ζ − )ζ + − iγ i ζ i + 12 iζ iζ i
. . (5.3.9)

In order to obtain the usual spinor field, it’s necessary to add a lagrange multiplier
term to the action constraining γ̂ = 0. This constraint can either be solved classically
(but only for even spacetime dimension D) by determining half of the ζ’s to be
the canonical conjugates of the other half (consider ζ 1 + iζ 2 vs. ζ 1 − iζ 2 , etc.), or
by imposing it quantum mechanically on the field Gupta-Bleuler style. The former
approach sacrifices manifest Lorentz invariance in the coordinate approach; however,
if the γ’s are considered simply as operators (without reference to their coordinate
representation), then the field is the usual spinor representation, and both can be
represented in the usual matrix notation. This constrained action is equivalent to the
second-order action

S = dτ dθ 12 G−1 (D 2 X) · (DX)

= dτ dθ 12 (GdX)d(GdX) ,


. . .
dτ ( 12 g −1x2 + ig −1 ψγ · x − 12 iγ · γ) , (5.3.10a)

or, in first-order form for x only,


. .
dτ (x · p − g 21 p2 + 12 iγ · γ + iψγ · p) . (5.3.10b)

The constraint γ ·p = 0 (the Dirac equation) is just a factorized form of the constraint
(2.2.8) for this particular representation of the Lorentz group.
A further constraint is necessary to get an irreducible Poincaré representation in
even D. Since any function of an anticommuting coordinate contains bosonic and
5.3. Spinning 89

fermionic terms as the coefficients of even and odd powers of that coordinate, we
need the constraint γ D = ±1 on the field (where γ D means just the product of all the
γ’s) to pick out a field of just one statistics (in this case, a Weyl spinor: notice that
D is even in order for the previous constraint to be applied). In the OSp approach
this Weyl chirality condition can also be obtained by an extension of the algebra
[4.10]: OSp(1,1|2)⊗U(1), where the U(1) is chiral transformations, results in an extra
Kronecker δ which is just the usual chirality projector. This U(1) generator (for at
least the special case of a Dirac spinor or Ramond string) can also be derived as a
constraint from first-quantization: The classical mechanics action for a Dirac spinor,
under the global transformation δγ a ∼
abc···d γ b γ c · · · γ d , varies by a boundary term

∼ dτ ∂τ

γ D , where as usual γ D ∼
abc···d γ a γ b γ c · · · γ d . By adding a lagrange multiplier
term for γ D ±1, this symmetry becomes a local one, gauged by the lagrange multiplier
(as for the other equations of motion). By 1D supersymmetrization, there is also a
lagrange multiplier for
abc···d pa γ b γ c · · · γ d . The action then describes a Weyl spinor.

Many supersymmetric gauges are possible for g and ψ. The simplest sets both
to constants (“temporal” gauge G = 1), but this gauge doesn’t allow an OSp(1,1|2)
algebra. The next simplest gauge, dG = 0, does the same to ψ but sets the τ
derivative of g to vanish, making it an extra coordinate in the field theory (related

to x− , or p+ ), giving the generators of (3.4.2). However, the gauge which also keeps
ψ as a coordinate (and as a partner to g) is G = 0. In order to get the maximal
coordinates (or at least zero-modes, for the string) we choose an OSp(1,1|2) which
keeps ψ (related to γ̃ ± , and the corresponding extra ghost, related to γ̃ α ). This gives
the modified BRST algebra of (3.5.1).

An “isospinning” particle [5.10] can be described similarly. By dropping the ψ
term in (5.3.6,10b) it’s possible to have a different symmetry on the indices of (γ, ζ)
than on those of (x, p). In fact, even the range of the indices and the metric can
be different. Thus, spin separates from orbital angular momentum and becomes
isospin. There is no longer an anticommuting gauge invariance, but with a positive
definite metric on the isospinor indices it’s no longer necessary to have one to maintain
unitarity. If we use the constraint γ̂ = 0 we get an isospinor, but if we don’t we get a
matrix, with the γ’s acting on one side and the γ̂’s on the other. Noting that τ reversal
switches γ with −γ̂, we see that the matrix gets transposed. Therefore, the complex
conjugation that is the quantum-mechanical analog of τ reversal is actually hermitian
conjugation, particularly on a field which is a matrix representation of some group.
(When γ̂ is constrained to vanish, τ reversal is not an invariance.) By combining

these anticommuting variables with the previous ones we get an isospinning spinning

At this point we take a slight diversion to discuss properties of spinors in arbitrary
dimensions with arbitrary spacetime signature. This will complete our discussion of
spinors in this section, and will be useful in the following section, where representa-
tions of supersymmetry, which is itself described by a spinor generator, will be found
to depend qualitatively on the dimension. The analysis of spinors in Euclidean space
(i.e, the usual spinor representations of SO(D)) can be obtained by the usual group
theoretical methods (see, e.g., [5.11]), using either Dynkin diagrams or an explicit
representation of the γ-matrices. The properties of spinors in SO(D + ,D − ) can then
be obtained by Wick rotation of D − space directions into time ones. (Of particu-
lar interest are the D-dimensional Lorentz group SO(D−1,1) and the D-dimensional
conformal group SO(D,2).) This affects the spinors with respect to only complex
conjugation properties. A useful notation to classify spinors and their properties
is: Denote a fundamental spinor (“spin 1/2”) as ψ α , and its hermitian conjugate
as −ψ̄ α . Denote another spinor ψ α such that the contraction ψ α ψ α is invariant un-
der the group, and its hermitian conjugate ψ̄ α . The representation of the group on
these various spinors is then related by taking complex conjugates and inverses of
the matrices representing the group on the fundamental one. For SO(D + ,D − ) there
are always some of these representations that are equivalent, since SO(2N) has only
2 inequivalent spinor representations and SO(2N+1) just 1. (In γ-matrix language,
the Dirac spinor can be reduced to 2 inequivalent Weyl spinors by projection with
2 (1 ± γ D ) in even dimensions.) In cases where there is another fundamental spinor
representation not included in this set, we also introduce a ψ α and the corresponding
3 other spinors. (However, in that case all 4 in each set will be equivalent, since there
are at most 2 inequivalent altogether.) Many properties of the spinor representations
can be described by classifying the index structure of: (1) the inequivalent spinors,
(2) the bispinor invariant tensors, or “metrics,” which are just the matrices relating
the equivalent spinors in the sets of 4, and (3) the σ-matrices (γ-matrices for D odd,
but in even D the matrices half as big which remain after Weyl projection), which are
simply the Clebsch-Gordan coefficients for relating spinor⊗spinor to vector. In the
latter 2 cases, we also classify the symmetry in the 2 spinor indices, where appropriate.

.The metrics are of 3 types (along with their complex conjugates and inverses): (1)
M , which gives charge conjugation for (pseudo)real representations, and is related

to complex conjugation properties of γ-matrices, (2) M . , which is the matrix which
5.3. Spinning 91

relates the Dirac spinors Ψ and Ψ̄, if it commutes with Weyl projection, and is related
to hermitian conjugation properties of γ-matrices, and (3) M αβ , which is the Clebsch-
Gordan coefficients for spinor⊗same-representation spinor to scalar, and is related
to transposition properties of γ-matrices. For all of these it’s important to know
whether the metric is symmetric or antisymmetric; in particular, for the first type
we get either real or pseudoreal representations, respectively. In γ-matrix language,
this charge conjugation matrix is straightforwardly constructed in the representation
where the γ-matrices are expressed in terms of direct products of the Pauli matrices
for the 2-dimensional subspaces. Upon Wick rotation of 1 direction each in any
number of pairs corresponding to these 2-dimensional subspaces, the corresponding
Pauli matrix factor in the charge conjugation matrix must be dropped (with perhaps
some change in the choice of Pauli matrix factors for the other subspaces). It then
follows that (pseudo)reality is the same in SO(D + +1,D − +1) as in SO(D + ,D − ), so
all cases follow from the Euclidean case. For the second type of metric, Ψ̄ = Ψ† in
the Euclidean case, so M . is just the identity matrix (i.e., the spinor representations
are unitary). After Wick rotation, this matrix becomes the product of all the γ-
matrices in the Wick rotated directions, since those γ-matrices got factors of i in the
Wick rotation, and thus need this extra factor to preserve the reality of the tensors
Ψ̄γ · · · γΨ. The symmetry properties of this metric then follow from those of the
γ-matrices. Also, because of the signature of the γ-matrices, it follows that this
metric, except in the Euclidean case, has half its eigenvalues +1 and half −1. The
last type of metric has only undotted indices and thus has nothing to do with complex
conjugation, so its properties are unchanged by Wick rotation. It’s identical to the
first type in Euclidean space (since the second type is the identity there; in general,
if 2 of the metrics exist, the third is just their product), which thus determines it in
the general case. Various types of groups are defined by these metrics alone (real,
unitary, orthogonal, symplectic, etc.), with the SO(D + ,D − ) group as a subgroup.
(In fact, these metrics completely determine the SO(D + ,D − ) group, up to abelian
factors, in D ≡ D + + D − ≤ 6, and allow all vector indices to be replaced by pairs
of spinor indices. They also determine the group in D = 8 for D − even, due to
“triality,” the discrete symmetry which permutes the vector representation with the
2 spinors.) We also classify the σ-matrices by their symmetry properties only when its
2 spinor indices are for equivalent representations, so they are unrelated to complex
conjugation (both indices undotted), and thus their symmetry is determined by the
Euclidean case.

We now summarize the results obtained by the methods sketched above for spinors
ψ, metrics η (symmetric) and Ω (antisymmetric), and σ-matrices, in terms of D mod
8 and D − mod 4:

D− 0 1 2 3
D Euclidean Lorentz conformal
ψ α ψ α
. ψα ψα . ψ α ψ α
. ψα ψα .
0 η αβ η α β η . η αβ η αβ Ωα β Ω . η αβ
. .
αβ αβ
σ αβ  σ σ αβ  σ
αβ αβ
. ψ
. α ψα
. ψ
. α

1 η αβ η α η β
. η αβ η α ηβ
. η αβ Ωα Ω β .
η αβ Ωα β Ω .
σ (αβ) σ (αβ) σ (αβ) σ (αβ)

. .
α α α
ψα ψ ψα ψ ψα ψ ψα ψα
2 η
. ηαβ Ω
. Ωα β
σ (αβ) σ (αβ) σ (αβ) σ (αβ) σ (αβ) σ (αβ) σ (αβ) σ (αβ)
. ψα
. ψα
. ψα
3 Ωαβ Ωα β η
. Ωαβ η α β Ω
. Ωαβ η α β Ω .
Ωαβ Ωα β η .
σ (αβ) σ (αβ) σ (αβ) σ (αβ)
ψ α ψ α
. ψα ψα . ψ α ψ α
. ψα ψα .
4 Ωαβ Ωα β η . Ωαβ Ωαβ η α β Ω . Ωαβ
. .
αβ αβ
σ αβ  σ σ αβ  σ
αβ αβ
. ψ
. α ψα
. ψ
. α

5 Ωαβ Ωα β η
. Ωαβ Ωα β η
. Ωαβ η α β Ω .
Ωαβ η α β Ω .
σ [αβ] σ [αβ] σ [αβ] σ [αβ]
ψα ψα ψα ψα
. ψα ψα ψα ψα
6 η
. Ωα β Ω
. ηαβ
[αβ] [αβ]
σ [αβ] σ σ [αβ] σ σ [αβ] σ [αβ] σ [αβ] σ [αβ]
. ψα
. ψα
. ψα
7 η αβ η α β η
. η αβ Ωα β Ω
. η αβ Ωα β Ω .
η αβ η α β η .
σ [αβ] σ [αβ] σ [αβ] σ [αβ]

(We have omitted the vector indices on the σ-matrices. We have also omitted metrics
which are complex conjugates or inverses of those shown, or are the same but with all
5.3. Spinning 93

indices primed, where relevant.) Also, not indicated in the table is the fact that η .
is positive definite for the Euclidean case and half-positive, half-negative otherwise.
Finally, the dimension of the spinors is 2(D−1)/2 for D odd and 2(D−2)/2 (Weyl spinor)
for D even. These N × N metrics define classical groups as subgroups of GL(N, C):

η αβ → SO(N, C)

Ωαβ → Sp(N, C)
η α β → GL(N, R)
Ωα β → GL*(N) (≡ U *(N))
. η
→ U(N) (or U( N2 , N2 ))

Ω . → U( N2 , N2 )
. αβ

η αβ η η .
→ SO(N) (or SO( N2 , N2 ))
. αβ

η αβ Ω Ω .
→ SO *(N)
. αβ

Ωαβ η Ω .
→ Sp(N)
. αβ

Ωαβ Ω η .
→ USp(N) (or USp( N2 , N2 ))

(When the matrix has a trace, the group can be factored into the corresponding
“S”-group times an abelian factor U(1) or GL(1, R).)
The σ-matrices satisfy the obvious relation analogous to the γ-matrix anticom-
mutation relations: Contract a pair of spinors on 2 σ-matrices and symmetrize in
the vector indices and you get (twice) the metric for the vector representation (the
SO(D + ,D − ) metric) times a Kronecker δ in the remaining spinor indices:

. . 
σ (a αα σ b)β α = σ (a αα σ b)βα = σ (a [αγ] σ b)[βγ] = σ (a (αγ) σ b)(βγ) = 2η ab δ α β , (5.3.11)

and similarly for expressions with dotted and undotted (or primed and unprimed)
indices switched. (We have raised indices with spinor metrics when necessary.)
Although (irreducible) spinors thus have many differences in different dimensions,
there are some properties which are dimension-independent, and it will prove useful
to change notation to emphasize those similarities. We therefore define spinors which
are real in all dimensions (or would be real after a complex similarity transformation,
and therefore satisfy a generalized Majorana condition). For those kinds of spinors
in the above table which are complex or pseudoreal, this means making a bigger

spinor which contains the real and imaginary components of the previous one as
independent components. If the original spinor was complex (D + −D − twice odd), the
new spinor is reducible to an irreducible spinor and its inequivalent complex conjugate
representation, which transform oppositely with respect to an internal U(1) generator
(“γ 5 ”). If the original spinor was pseudoreal (D + − D − = 3, 4, 5 mod 8), the new
spinor reduces to 2 equivalent irreducible spinor representations, which transform as
a doublet with respect to an internal SU(2).

The net result for these real spinors is that we have the following analog of the
above table for those properties which hold for all values of D + :

D− 0 1 2 3
Euclidean Lorentz conformal
ψ α ψ α ψα ψα ψ α ψ α ψα ψα
η αβ Ωαβ
γ αβ  γ (αβ) γ (αβ) γ αβ  γ [αβ] γ [αβ]

These γ-matrices satisfy the same relations as the σ-matrices in (5.3.11). (In fact, they
are identical for D + = D − mod 8.) Their additional, D + -dependent properties can
be described by additional metrics: (1) the internal symmetry generators mentioned
above; and (2) for D odd, a metric M αβ or M αβ  which relates the 2 types of spinors
(since there are 2 independent irreducible spinor representations only for D even).

Similar methods of first-quantization will be applied in sect. 7.2 to the spinning
string, which has spin-0 and spin-1/2 ground states. Classical mechanics actions for
particles with other spins (or strings with ground states with other spins), i.e., gauge
fields, are not known. (For the superstring, however, a nonmanifestly supersymmetric
formalism can be obtained by a truncation of the spinning string, eliminating some
of the ground states.) On the other hand, the BRST approach of chap. 3 allows
the treatment of the quantum mechanics of arbitrary gauge fields. Furthermore, the
superparticle, described in the following section, is described classical-mechanically
by a spin-0 or spin-1/2 superfield, which includes component gauge fields, just as the
string has component gauge fields in its excited modes.
5.4. Supersymmetric 95

5.4. Supersymmetric

The superparticle is obtained from the spinless particle by symmetrizing with
respect to the supersymmetry of the higher-dimensional space in which the one-
dimensional world line of the particle is imbedded. (For reviews of supersymmetry,
see [1.17].) As for the spinless particle, a full understanding of this action consists of
just understanding the algebras of the covariant derivatives and equations of motion.
In order to describe arbitrary D, we work with the general real spinors of the pre-
vious section. The covariant derivatives are pa (momentum) and dα (anticommuting
spinor), with
{dα , dβ } = 2γ a αβ pa (5.4.1)

(the other graded commutators vanish), where the γ matrices are symmetric in their
spinor indices and satisfy
γ (a αγ γ b)βγ = 2η ab δ α β , (5.4.2)

as described in the previous section. This algebra is represented in terms of coordi-
nates xa (spacetime) and θα (anticommuting), and their partial derivatives ∂ a and
∂ α , as
pa = i∂ a , dα = ∂ α + iγ a αβ θβ ∂ a . (5.4.3)

These covariant derivatives are invariant under supersymmetry transformations gen-
erated by pa and q α , which form the algebra

{q α , q β } = −2γ a αβ pa . (5.4.4)

pa is given above, and q α is represented in terms of the same coordinates as

q α = ∂ α − iγ a αβ θβ ∂ a . (5.4.5)

(See (5.3.1) for D = 1.) All these objects also transform covariantly under Lorentz
transformations generated by

J ab = −ix[a pb] + 14 θα γ [aαγ γ b] γβ ∂ β + M ab , (5.4.6)

where we have included the (coordinate-independent) spin term M ab . (In comparison
with (2.2.4), the spin operator here gives just the spin of the superfield, which is
a function of x and θ, whereas the spin operator there includes the θ∂ term, and
thus gives the spin of the component fields resulting as the coefficients of a Taylor
expansion of the superfield in powers of θ.)

As described in the previous section, even for “simple” supersymmetry (the small-
est supersymmetry for that dimension), these spinors are reducible if the irreducible
spinor representation isn’t real, and reduce to the direct sum of an irreducible spinor
and its complex conjugate. However, we can further generalize by letting the spinor
represent more than one of such real spinors (and some of each of the 2 types that
are independent when D is twice odd), and still use the same notation, with a single
index representing all spinor components. (5.4.1-6) are then unchanged (except for
the range of the spinor index). However, the nature of the supersymmetry repre-
sentations will depend on D, and on the number of minimal supersymmetries. In
the remainder of this section we’ll stick to this notation to manifest those properties
which are independent of dimension, and include such things as internal-symmetry
generators when required for dimension-dependent properties.

For a massless, real, scalar field, p2 = 0 is the only equation of motion, but for a
massless, real, scalar superfield, the additional equation p/d = 0 (where d is the spinor
derivative) is necessary to impose that the superfield is a unitary representation of
(on-shell) supersymmetry [5.12]: Since the hermitian supersymmetry generators q
pq, p/q} ∼ p2 p/ = 0, but on unitary representations
satisfy {q, q} ∼ p, we have that {/
any hermitian operator whose square vanishes must also vanish, so 0 = p/q = p/d up to
a term proportional to p2 = 0. This means that only half the q’s are nonvanishing. We
can further divide these remaining q’s in (complex) halves as creation and annihilation
operators. A massless, irreducible representation of supersymmetry is then specified
in this nonmanifestly Lorentz covariant approach by fixing the “Clifford” vacuum of
these creation and annihilation operators to be an irreducible representation of the
Poincaré group.

Unfortunately, the p2 and p/d equations are not sufficient to determine an irre-
ducible representation of supersymmetry, even for a scalar superfield (with certain
exceptions in D ≤ 4) since, although they kill the unphysical half of the q’s, they
don’t restrict the Clifford vacuum. The latter restriction requires extra constraints
in a manifestly Lorentz covariant formalism. There are several ways to find these
additional constraints: One is to consider coupling to external fields. The simplest
case is external super-Yang-Mills (which will be particularly relevant for strings). The
generalization of the covariant derivatives is

dα → ∇α = dα + Γα ,

pa → ∇a = pa + Γa . (5.4.7)
5.4. Supersymmetric 97

We thus have a graded covariant derivative ∇A , A = (a, α). Without loss of generality,
we consider cases where the only physical fields in the super-Yang-Mills multiplet are
a vector and a spinor. The other cases (containing scalars) can be obtained easily by
dimensional reduction. Then the commutation relations of the covariant derivatives
become [5.13]
{∇α , ∇β } = 2γ a αβ ∇a ,

[∇α , ∇a ] = 2γ aαβ W β ,

[∇a , ∇b ] = F ab , (5.4.8)
where W α is the super-Yang-Mills field strength (at θ = 0, the physical spinor field),
and consistency of the Jacobi (Bianchi) identities requires
γ a(αβ γ a γ)δ = 0 . (5.4.9)
This condition (when maximal Lorentz invariance is assumed, i.e., SO(D-1,1) for a
taking D values) implies spacetime dimensions D = 3, 4, 6, 10 , and “antispacetime”
dimensions (the number of values of the index α) D  = 2(D − 2). (The latter identity
follows from multiplying (5.4.9) by γ bαβ and using (5.4.2).) The generalization of the
equations of motion is [5.14]
p/d → γ aαβ ∇a ∇β ,
1 2
2p → 12 ∇a ∇a + W α ∇α , (5.4.10)
but closure of this algebra also requires new equations of motion which are certain
Lorentz pieces of ∇[α ∇β] . Specifically, in D = 3 there is only one Lorentz piece (a
scalar), and it gives the usual field equations for a scalar multiplet [5.15]; in D = 4 the
scalar piece again gives the usual equations for a chiral scalar multiplet, but the (axial)
vector piece gives the chirality condition (after appropriate normal ordering); in D = 6
only the self-dual third-rank antisymmetric tensor piece appears in the algebra, and
it gives equations satisfied by scalar multiplets (but not by tensor multiplets, which
are also described by scalar superfield strengths, but can’t couple minimally to Yang-
Mills) [5.16]; but in D = 10 no multiplet is described because the only one possible
would be Yang-Mills itself, but its field strength W α carries a Lorentz index, and the
equations described above (which apply only to scalars) need extra terms containing
Lorentz generators.
Another way to derive the modifications is to use superconformal transformations.
The superconformal groups [5.17] are actually easier to derive than the supersymme-
try groups because they are just graded versions of classical groups. Specifically, the

classical supergroups (see [5.18] for a review) have defining representations defined in
terms of a metric M
. , which makes them unitary (or pseudounitary, if the metric
isn’t positive definite), and sometimes also a graded-symmetric metric M AB , and thus
M A B by combining them (and their inverses). The generators which have bosonic-
bosonic or fermionic-fermionic indices are bosonic, and those with bosonic-fermionic
are fermionic. (The choice of which parts of the A index are bosonic and which are
fermionic can be reversed, but this doesn’t affect the statistics of the group genera-
tors.) Since the bosonic subgroup of the supergroups with just the M . metric is
the direct product of 2 unitary groups, those supergroups are called (S)SU(M|N) for
M values of the index of one statistics and N of the other, where the S is because a
(graded) trace condition is imposed, and there is a second S for M=N because then
a second trace can be removed (so each of the 2 unitary subgroups becomes SU(N)).
The supergroups which also have the graded-symmetric metric M AB have a bosonic
subgroup which is orthogonal in the sector where the metric is symmetric and sym-

plectic in the sector where it is antisymmetric. In this case we choose the metric M .
also to have graded symmetry, in such a way that the metric M A B obtained from
their product is totally symmetric, so the defining representation is real, or totally
antisymmetric, so the representation is pseudoreal. The former is generally called
OSp(M|2N), and we call the latter OSp*(M|2N).
We next assume that the anticommuting generators of these supergroups are
to be identified with the conformal generalization of the supersymmetry generators.
Thus, one index is to be identified with an internal symmetry, and the other with a
conformal spinor index. The conformal spinor reduces to 2 Lorentz spinors, one of
which is the usual supersymmetry, the “square root” of translations, and the other
of which is “S-supersymmetry,” the square root of conformal boosts. The choice of
supergroup then follows immediately from the graded generalization of the conformal
spinor metrics appearing in the table of the previous section [5.19]:

D mod 8 superconformal bosonic subgroup dim-0/dilatations
0,4 (S)SU(N|ν,ν) SU(ν,ν)⊗SU(N)(⊗U(1)) SL(ν,C)⊗(S)U(N)
1,3 OSp(N|2ν) Sp(2ν)⊗SO(N) SL(ν,R)⊗SO(N)
2 OSp(N|2ν) ” ”
or (”)+ ⊗(”)− (”)+ ⊗(”)− (”)+ ⊗(”)−
5,7 OSp*(2ν|2N) SO*(2ν)⊗USp(2N) SL*(ν)⊗USp(2N)
6 OSp*(2ν|2N) ” ”
or (”)+ ⊗(”)− (”)+ ⊗(”)− (”)+ ⊗(”)−
5.4. Supersymmetric 99

where “dim-0” are the generators which commute with dilatations (see sect. 2.2), so
the last column gives Lorentz⊗internal symmetry (at least). ν is the dimension of
the (irreducible) Lorentz spinor (1/2 that of the conformal spinor), N is the number
of minimal supersymmetries, and D(>2) is the dimension of (Minkowski) spacetime
(with conformal group SO(D,2), so 2 less than the D in the previous table). The 2
choices for twice-odd D depend on whether we choose to represent the superconformal
group on both primed and unprimed spinors. If so, there can be a separate N and
N . Again, we have used * to indicate groups which are Wick rotations such that
the defining representation is pseudoreal instead of real. (SL* is sometimes denoted
Unfortunately, as discussed in the previous section, the bosonic subgroup acting
on the conformal spinor part of the defining representation, as defined by the spinor
metrics (plus the trace condition, when relevant) gives a group bigger than the con-
formal group unless D ≤ 6. However, we can still use D ≤ 6, and perhaps some of the
qualitative features of D > 6, for our analysis of massless field equations. We then
generalize our analysis of sect. 2.2 from conformal to superconformal. It’s sufficient
to apply just the S-supersymmetry generators to just the Klein-Gordon operator. We
then find [2.6,5.19]:
1 b

1 2
2 , J ab } + 12 {pa , ∆} = pb M ab + pa d − D−2
2p → p/q = p/d → , (5.4.11)
q [α q β] + · · · = d[α dβ] + · · ·
where the last expression means certain Lorentz pieces of dd plus certain terms con-
1 αβ
taining Lorentz and internal symmetry generators. In particular, 16 γ abc dα dβ +
2 p[a M bc] is the supersymmetric analog of the Pauli-Lubansky vector [5.20]. The vec-
tor equation is (2.2.8) again, derived in essentially the same way.
For the constraints we therefore choose [5.21]

A = 12 p2 ,

Bα = γ aαβ pa dβ ,
C abc = αβ
16 γ abc dα dβ + 12 p[a M bc] ,

D a = M a b pb + kpa ; (5.4.12a)

or, in matrix notation,

A = 12 p2 ,

B = p/d ,

C abc = 16 dγ abc d + 12 p[a M bc] ,

D a = M a b pb + kpa ; (5.4.12b)

where out of (5.4.11) we have chosen A and D as for nonsupersymmetric theories
(sect. 2.2), B for unitarity (as explained above), and just the Pauli-Lubansky part
of the rest (which is all of it for D=10), the significance of which will be explained

These constraints satisfy the algebra

{B, B} = 4/
pA ,

[D a , D b ] = −2M ab A − p[a D b] ,

[C abc , B] = −8γ abc dA ,

[C abc , D d ] = η d[a pb D c] ,

[C abc , C def ] = − 14 [δ [d [a pe C bc] f ] − (abc ↔ def )] − 1
128 d(4γ
def − 2δ [d [a δ e b γ c] f ] )B ,

rest = 0 , (5.4.13)

with some ambiguity in how the right-hand side is expressed due to the relations

p · D = 2kA ,

p/B = 2dA ,

dB = 2(tr I)A ,
1 1
6 p[a C bcd] = − 16 dγ abcd B ,
pc C cab = 2M ab A + p[a D b] + 16 dγ ab B . (5.4.14)

(γ ab = 12 γ [a γ b] , etc.)

In the case of supersymmetry with an internal symmetry group (extended su-
persymmetry, or even simple supersymmetry in D=5,6,7), there is an additional con-
straint analogous to C abc for superisospin:

C a,int = 18 dγ a σ int d + pa M int . (5.4.15)

σ int are the matrix generators of the internal symmetry group, in the representation
to which d belongs, and M int are those which act on the external indices of the
5.4. Supersymmetric 101

Unfortunately, there are few superspin-0 multiplets that are contained within
spin-0, isospin-0 superfields (i.e., that themselves contain spin-0, isospin-0 compo-
nent fields). In fact, the only such multiplets of physical interest in D>4 are N=1
Yang-Mills in D=9 and N=2 nonchiral supergravity in D=10. (For a convenient listing
of multiplets, see [5.22].) However, by the method described in the previous section,
spinor representations for the Lorentz group can be introduced. By including “γ-
matrices” for internal symmetry, we can also introduce defining representations for the
internal symmetry groups for which they are equivalent to the spinor representations
of orthogonal groups (i.e., SU(2)=USp(2)=SO(3), USp(4)=SO(5), SU(4)=SO(6),
SO(4)-vector=SO(3)-spinor⊗SO(3) -spinor, SO(8)-vector=SO(8) -spinor). Further-
more, arbitrary U(1) representations can be described by adding extra terms without
introducing additional coordinates. This allows the description of most superspin-
0 multiplets, but with some notable exceptions (e.g., 11D supergravity). However,
these equations are not easily generalized to nonzero superspins, since, although the
superspin operator is easy to identify in the light-cone formalism (see below), the cor-
responding operator would be nonlocal in a covariant description (or appear always
with an additional factor of momentum).
We next consider the construction of mechanics actions. These equations describe
only multiplets of superspin 0, i.e., the smallest representations of a given supersym-
metry algebra, for reasons to be described below. (This is no restriction in D=3,
where superspin doesn’t exist, and in D=4 arbitrary superspin can be treated by a
minor modification, since there superspin is abelian.) As described in the previous
section, only spin-0 and spin-1/2 superfields can be described by classical mechanics,
and we begin with spin-0, dropping spin terms in (5.4.12), and the generator D. The
action is then given by (3.1.10), where [5.23]

z M = (xm , θµ ) , π A = (pa , idα ) ,
. . . .
z M eM A (z) = (x − iθγθ, θ) ,

iG i (π) = (A, Bα , C abc ) . (5.4.16)

Upon quantization, the covariant derivatives become

π A = ieA M ∂ M = i(∂ a , ∂ α + iγ a αβ θβ ∂ a ) , (5.4.17)

which are invariant under the supersymmetry transformations

δx = ξ − i
γθ , δθ =
. (5.4.18)

The transformation laws then follow directly from (3.1.11), with the aid of (5.4.13)
for the λ transformations.
The classical mechanics action can be quantized covariantly by BRST methods.
In particular, the transformations generated by B [5.24] (with parameter κ) close on
those generated by A (with parameter ξ):

δx = ξp + iκ(γd + p/γθ) , δθ = p/κ ,

δp = 0 , δd = 2p2 κ ,
δg = ξ + 4iκ/
pψ ,
δψ = κ . (5.4.19)
Because of the second line of (5.4.14), the ghosts have a gauge invariance similar to
the original κ invariance, and then the ghosts of those ghosts again have such an
invariance, etc., ad infinitum. This is a consequence of the fact that only half of θ
can be gauged away, but there is generally no Lorentz representation with half the
components of a spinor, so the spinor gauge parameter must itself be half gauge, etc.
Although somewhat awkward, the infinite set of ghosts is straightforward to find.
Furthermore, if derived from the light cone, the OSp(1,1|2) generators automatically
contain this infinite number of spinors: There, θ is first-quantized in the same way as
the Dirac spinor was second-quantized in sect. 3.5, and θ obtains an infinite number
of components (as an infinite number of ordinary spinors) as a result of being a
representation of a graded Clifford algebra (specifically, the Heisenberg algebras of
γ α and γ̃ α ). This analysis will be made in the next section.
On the other hand, the analysis of the constraints is simplest in the light-cone
formalism. The A, B, and D equations can be solved directly, because they are all of
the form p · f = p+ f − + · · · :
1 2
A=0 → p− = − pi ,
B=0 → γ −d = − γ i pi γ − γ + d ,
D=0 → M −i = (M i j pj + kpi ) , M −+ = k , (5.4.20a)
where we have chosen the corresponding gauges

x+ = 0 ,

γ +θ = 0 ,

M +i = 0 . (5.4.20b)
5.4. Supersymmetric 103

These solutions restrict the x’s, θ’s, and Lorentz indices, respectively, to those of the
light cone. (Effectively, D is reduced by 2, except that p+ remains.)

However, a superfield which is a function of a light-cone θ is not an irreducible
representation of supersymmetry (except sometimes in D≤4), although it is a unitary
one. In fact, C is just the superspin operator which separates the representations: Due
to the other constraints, all its components are linearly related to

C +ij = p+ M ij + 16 dγ + γ ij d . (5.4.21)

Up to a factor of p+ , this is the light-cone superspin: On an irreducible representation
of supersymmetry, it acts as an irreducible representation of SO(D-2). In D=4 this
can be seen easily by noting that the irreducible representations can be represented
in terms of chiral superfields (d¯ = 0) with different numbers of d’s acting on them,
and the dd¯ in C just counts the numbers of d’s. In general, if we note that the full
light-cone Lorentz generator can be written as

J ij = −ix[i pj] + 12 θγ ij + M ij
1 1
= −ix[i pj] − qγ + γ ij q + dγ + γ ij d + M ij
16p+ 16p+
= Jˆij + C +ij , (5.4.22)

then, by expressing any state in terms of q’s acting on the Clifford vacuum, we see that
Jˆ gives the correct transformation for those q’s and the x-dependence of the Clifford
vacuum, so C/p+ gives the spin of the Clifford vacuum less the contribution of the
qq term on it, i.e., the superspin. Unfortunately, the mechanics action can’t handle
spin operators for irreducible representations (either for M ij or the superspin), so we
must restrict ourselves not only to spin 0 (referring to the external indices on the
superfield), but also superspin 0 (at least at the classical mechanics level). Thus, the
remaining constraint C = 0 is the only possible (first-class) constraint which can make
the supersymmetry representation irreducible. The constraints (5.4.12) are therefore
necessary and sufficient for deriving the mechanics action. However, if we allow the
trivial kind of second-class constraints that can be solved in terms of matrices, we can
generalize to spin-1/2. In principle, we could also do superspin-1/2, but this leads to
covariant fields which are just those for superspin-0 with an extra spinor index tagged
on, which differs by factors of momentum (with appropriate index contractions) from
the desired expressions. (Thus, the superspin operator would be nonlocal on the

latter.) Isospin-1/2 can be treated similarly (and superisospin-1/2, but again as a
tagged-on index).

The simplest nontrivial example is D=4. (In D=3, C +ij vanishes, since the trans-
verse index i takes only one value.) There light-cone spinors have only 1 (complex)
component, and so does C +ij . For this case, we can (and must, for an odd number N
of supersymmetries) modify C abc :

C abc = 16 dγ abc d + 12 p[a M bc] + iH
abcd pd , (5.4.23)

where H is the “superhelicity.” We then find
C +ij = p+
ij M + iH − i [da , d¯a ] , (5.4.24)

where M ij =
ij M and {da , d¯b } = p+ δ a b , and a is an SU(N) index. The “helicity”
h is given by M = −ih, and we then find by expanding the field over chiral fields φ
¯ = 0)
[5.25-27] (dφ

ψ ∼ (d)n φ → H = h + 14 (2n − N) . (5.4.25)

Specifying both the spin and superhelicity of the original superfield fixes both h and
H, and thus determines n. Note that this requires H to be quarter-(odd-)integral for
odd N. In general, the SU(N) representation of φ also needs to be specified, and the
relevant part of (5.4.15) is

C +a = p+ M a
b b
−i [da , d¯b ] − 14 δ a b [dc , d¯c ] , (5.4.26)

and the vanishing of this quantity forces φ to be an SU(N)-singlet. (More general
cases can be obtained simply by tacking extra indices onto the original superfield,
and thus onto φ.)

We next consider 10D super Yang-Mills. The appropriate superfield is a Weyl
or Majorana spinor, so we include terms as in the previous section in the mechanics
action. To solve the remaining constraint, we first decompose SO(9,1) covariant
spinors and γ-matrices to SO(8) light-cone ones as
dα = 2 ,
√    √ 
2p+ p/T Σ 0 Σ 0 2p− p/T
p/αβ = √ , p/ αβ
= √ ,
p/T † − 2p− 0 Σ 0 Σ p/T † − 2p+
5.4. Supersymmetric 105

a/T b/T † + b/T a/T † = a/T † b/T + b/T † a/T = 2aT · bT ,

p/T * = Σ/
pT Σ , Σ = Σ† = Σ* , Σ2 = I . (5.4.27)

(We could choose the Majorana representation Σ = I, but other representations can
be more convenient.) The independent supersymmetry-covariant derivatives are then

d+ = +
+ p+ Σθ+ , pT , p+ . (5.4.28)
In order to introduce chiral light-cone superfields, we further reduce SO(8) to
SO(6)⊗SO(2)=U(4) notation:
√ da pab δ a b pL 0 I
d+ = 2 ¯a , p/T = , Σ=
d δ b a p̄L p̄ab I 0

(p̄ab = 12
abcd pcd ) . (5.4.29)

In terms of this “euphoric” notation, the constraints C +ij are written on the SO(6)-
spinor superfield as
1 ψc δ c b ψ a − 14 δ a b ψ c 1
C +a b = i 12 −i ¯ 1 ¯
[da , d ] − 4 δ a [dd , d ]
b b d
p+ ψ̄ c −δ a c ψ̄ b + 14 δ a b ψ̄ c 4p+ ψ̄ c
1 ψc 1 −ψ c ψc
C+ c
= c
i 12
−i [da , d¯a ] ,
p+ ψ̄ ψ̄ 4p+ ψ̄ c
1 ψc 1

abcd ψ̄ d 1 ψc
C +ab = i2 −i da db , (5.4.30)
p+ ψ̄ c 0 2p+ ψ̄ c
and the complex conjugate equation for C + ab . (Note that it is crucial that the original
SO(10) superfield ψ α was a spinor of chirality opposite to that of dα in order to obtain
soluble equations.) The solution to the first 2 equations gives ψ in terms of a chiral
superfield φ,
ψ a = da φ , (5.4.31a)

and that to the third equation imposes the self-duality condition [5.25-27]

1 abcd
da db dc dd φ = p+ 2 φ̄ . (5.4.31b)

This can also be written as
−1/2 aπ
2 p+ d φ = φ̄ → dθ eθ a p+ /2
φ(θa ) = [φ(π̄ a )] * . (5.4.32)

This corresponds to the fact that in the mechanics action τ reversal on θα includes
multiplication by the charge conjugation matrix, which switches θa with θ̄a , which
equals −(2/p+ )∂/∂θa by the chirality condition d¯a = 0.

These results are equivalent to those obtained from first-quantization of a me-
chanics action with dα = 0 as a second-class constraint [5.28]. (This is the analog of
the constraint γ̂ = 0 of the previous section.) This is effectively the same as drop-
ping the d terms from the action, which can then be written in second-order form by
eliminating p by its equation of motion. This can be solved either by using a chiral
superfield [5.27] as a solution to this constraint in a Gupta-Bleuler formalism,

d¯a φ = 0 → da φ̄ = 0 → φ̄d+ φ = 0 , (5.4.33)

or by using a superfield with a real 4-component θ [5.26] as a solution to this con-
straint before quantization (but after going to a light-cone gauge), determining half
the remaining components of γ − θ to be the canonical conjugates of the other half.
However, whereas either of these methods with second-class constraints requires the
breaking of manifest Lorentz covariance just for the formulation of the (field) theory,
the method we have described above has constraints on the fields which are mani-
festly Lorentz covariant ((5.4.12)). Furthermore, this second-class approach requires
that (5.4.32) be imposed in addition, whereas in the first-class approach it and the
chirality condition automatically followed together from (5.4.21) (and the ordinary
reality of the original SO(9,1) spinor superfield).
On the other hand, the formalism with second-class constraints can be derived
from the first-class formalism without M ab terms (and thus without D in (5.4.12))
[5.29]: Just as A and B were solved at the classical level to obtain (5.4.20), C = 0 can
also be solved classically. To be specific, we again consider D = 10. Then C +ij = 0
is equivalent to d[µ dν] = 0 (where µ is an 8-valued light-cone spinor index). (They
are just different linear combinations of the same 28 antisymmetric quadratics in
d, which are the only nonvanishing d products classically.) This constraint implies
the components of d are all proportional to the same anticommuting scalar, times
different commuting factors:

dµ dν = 0 → dµ = cζ µ , (5.4.34a)

where c is anticommuting and ζ commuting. Furthermore, C +ij are just SO(8) gen-
erators on d, and thus their gauge transformation can be used to rotate it in any
direction, thus eliminating all but 1 component [5.30]. (This is clear from triality,
since the spinor 8 representation is like the vector 8.) Specifically, we choose the
gauge parameters of the C transformations to depend on ζ in such a way as to rotate
ζ in any one direction, and then redefine c to absorb the remaining ζ factor:

C gauge : dµ = δ m 1 c . (5.4.34b)
5.4. Supersymmetric 107

In this gauge, the C constraint itself is trivial, since it is antisymmetric in d’s. Finally,
we quantize this one remaining component c of d to find

quantization : c2 = p+ → c = ± p+ . (5.4.34c)

c has been determined only up to a sign, but there is a residual C gauge invariance,
since the C rotation can also be used to rotate ζ in the opposite direction, changing its
sign. After using the gauge invariance to make all but one component of ζ vanish, this
sign change is the only part of the gauge transformation which survives. It can then
be used to choose the sign in (5.4.34c). Thus, all the d’s are determined (although 1
component is nonvanishing), and we obtain the same set of coordinates (x and q, no
d) as in the second-class formalism. The C constraint can also be solved completely
at the quantum mechanical level by Gupta-Bleuler methods [5.29]. The SO(D-2)
generators represented by C are then divided up into the Cartan subalgebra, raising
operators, and lowering operators. The raising operators are imposed as constraints
(on the ket, and the lowering operators on the bra), implying only the highest-weight
state survives, and the generators of the Cartan subalgebra are imposed only up to
“normal-ordering” constants, which are just the weights of that state.
The components of this chiral superfield can be identified with the usual vector
+ spinor [5.26,27]:
φ(x, θa ) =
(p+ )−1 AL (x) + θa (p+ )−1 χa (x) + θ2ab Aab (x) + θ3 a χ̄a (x) + θ4 (p+ )ĀL (x) , (5.4.35)
and the θ = 0 components of γ + ψ = (ψ a , ψ̄ a ) can be identified with the spinor. Alter-
natively, the vector + spinor content can be obtained directly from the vanishing of
C +ij of (5.4.21), without using euphoric notation: We first note that the Γ-matrices of
M ij are represented by 2 spinors, corresponding to the 2 different chiralities of spinors
in SO(8). SO(8) has the property of “triality,” which is the permutation symmetry
of these 2 spinors with the vector representation. (All are 8-component representa-
tions.) Since the anticommutation relations of d are just a triality transformation of
those of the Γ’s (modulo p+ ’s), they are represented by the other 2 representations:
a spinor of the other chirality and a vector. The same holds for the representation
of q. Thus, the direct product of the representations of Γ and d includes a singlet
(superspin 0), picked out by C +ij = 0, so the total SO(8) representation (generated
by J ij of (5.4.22)) is just that of q, a spinor (of opposite chirality) and a vector.
There is another on-shell method of analysis of (super)conformal theories that is
manifestly covariant and makes essential use of spinors. This method expresses the

fields in terms of the “spinor” representation of the superconformal group. (The ordi-
nary conformal group is the case N=0.) The spinor is defined in terms of generalized
γ-matrices (“twistors” [5.31] or “supertwistors” [5.32]):

{γ̄ A , γ B ] = δ A B , (5.4.36)

where the index has been lowered by M
. , and the grading is such that the conformal
spinor part has been chosen commuting and the internal part anticommuting, just as
ordinary γ-matrices have bosonic (vector) indices but are anticommuting. The anti-
commuting γ’s are then closely analogous to the light-cone supersymmetry generators
γ − q. The generators are then represented as

GA B ∼ γ A γ̄ B , (5.4.37)

with graded (anti)symmetrization or traces subtracted, as appropriate. The case
of OSp(1,1|2) has been treated in (3.5.1). A representation in terms of the usual
superspace coordinates can then be generated by coset-space methods, as described in
sect. 2.2. We begin by identifying the subgroup of the supergroup which corresponds
to supersymmetry (by picking 1 of the 2 Lorentz spinors in the conformal spinor
generator) and translations (by closure of supersymmetry). We then equate their
ˆ of sect. 2.2) with their representation
representation in (5.4.37) (analogous to the J’s
in (5.4.3,5) (analogous to the J’s of sect. 2.2). (The constant γ-matrices of (5.4.3,5)
should not be confused with the operators of (5.4.36).) This results in an expression
analogous to (2.2.6), where Φ(0) is a function of half of (linear combinations of)
the γ’s of (5.4.36) (the other half being their canonical conjugates). (For example,
the bosonic part of γ A is a conformal spinor which is expressed as a Lorentz spinor
ζ α and its canonical conjugate. (5.4.37) then gives pa = γ a αβ ζ α ζ β , which implies
p2 = 0 in D = 3, 4, 6, 10 due to (5.4.9).) We then integrate over these γ’s to obtain
a function of just the usual superspace coordinates x and θ. Due to the quadratic
form of the momentum generator in terms of the γ’s, it describes only positive energy.
Negative energies, for antiparticles, can be introduced by adding to the field a term
for the complex conjugate representation. At least in D=3,4,6,10 this superfield
satisfies p2 = 0 as a consequence of the explicit form of the generators (5.4.37), and
as a consequence all the equations which follow from superconformal transformations.
These equations form a superconformal tensor which can be written covariantly as an
expression quadratic in GA B . In D=4 an additional U(1) acting on the twistor space
can be identified as the (little group) helicity, and in D=6 a similar SU(2)(⊗SU(2) if
5.4. Supersymmetric 109

the primed supergroup is also introduced) appears. ((5.4.34a) is also a supertwistor
type of relation.)
The cases D=3,4,6,10 [5.33] are especially interesting not only for the above
reason and (5.4.9) but also because their various spacetime groups form an interesting
pattern if we consider these groups to be the same for these different dimensions
except that they are over different generalized number systems A called “division
algebras.” These are generalizations of complex numbers which can be written as
z = z0 + 1 z i ei , {ei , ej } = −2δ ij , where n=0,1,3, or 7. Choosing for the different
dimensions the division algebras

3 real
4 complex
6 quaternion
10 octonion

we have the correspondence

SL1 (1, A) = SO(D − 2) 

SU(2, A) = SO(D − 1) 

SU(1, 1, A) = SO(D − 2, 1)
/SO(D − 3)
SL(2, A) = SO(D − 1, 1) 

SU  (4, A) = SO(D, 2) 

SU(N|4, A) = superconf ormal
where SL1 means only the real part (z 0 ) of the trace of the defining representation
vanishes, by SU  we mean traceless and having the metric Ω . (vs. η . for SU),
. = (η ., Ω .) (in that order). The . refers
αβ αβ
and the graded SU has metric M
AB ab αβ
to generalized conjugation ei → −ei (and the ei are invariant under transposition,
although their ordering inside the matrices changes). The “/SO(D − 3)” refers to
the fact that to get the desired groups we must include rotations of the D − 3 ei ’s
among themselves. The only possible exception is for the D = 10 superconformal
groups, which don’t correspond to the OSp(N|32) above, and haven’t been shown
to exist [5.34]. The light-cone form of the identity (5.4.9) ((7.3.17)) is equivalent to
the division algebra identity |xy| = |x||y| (|x|2 = xx* = x0 2 + xi 2 ), where both
the vector and spinor indices on the light-cone γ-matrices correspond to the index for
(z 0 , z i ) (all ranging over D − 2 values) [5.35].

A similar first-quantization analysis will be made for the superstring in sect. 7.3.

5.5. SuperBRST

Instead of using the covariant quantization which would follow directly from the
constraint analysis of (5.4.12), we will derive here the BRST algebra which follows
from the light-cone by the method of sect. 3.6, which treats bosons and fermions sym-
metrically [3.16]. We begin with any (reducible) light-cone Poincaré representation
which is also a supersymmetry representation, and extend also the light-cone super-
symmetry generators to 4+4 extra dimensions. The resulting OSp(D+1,3|4) spinor
does not commute with the BRST OSp(1,1|2) generators, and thus mixes physical and
unphysical states. Fortunately, this extended supersymmetry operator q can easily
be projected down to its OSp(1,1|2) singlet piece q 0 . We begin with the fact that the
light-cone supersymmetry generator is a tensor operator in a spinor representation of
the Lorentz group:
[J ab , q] = − 12 γ ab q , (5.5.1)

where γ ab = 12 γ [a γ b] . (All γ’s are now Dirac γ-matrices, not the generalized γ’s of
(5.4.1,2).) As a result, its extension to 4+4 extra dimensions transforms with respect
to the U(1)-type OSp(1,1|2) as

[J AB , q} = − 12 (γ AB + γ A B )q . (5.5.2)

It will be useful to combine γ A and γ A into creation and annihilation operators as in

γ A = aA + a† A , γ A = i(aA − a† A ) ; {aA , a† B ] = η AB (5.5.3)

→ 1
2 (γ AB + γ A B ) = a† [A aB) . (5.5.4)

(a† A aB , without symmetrization, are a representation of U(1, 1|1, 1).) We choose
boundary conditions such that all “states” can be created by the creation operators
a† from a “vacuum” annihilated by the annihilation operators a. (This choice, elim-
inating states obtained from a second vacuum annihilated by a† , is a type of Weyl
projection.) This vacuum is a fermionic spinor (acted on by γ a ) whose statistics are
changed by a†α (but not by a† ± ). If q is a real spinor, we can preserve this reality by

choosing a representation where γ A is real and γ A is imaginary. (In the same way, for
the ordinary harmonic oscillator the ground state can be chosen to be a real function
5.5. SuperBRST 111

of x, and the creation operator ∼ x − ∂/∂x preserves the reality.) The corresponding
charge-conjugation matrix is C = iγ 5 , where
1 c ,γ c̃ }
γ 5 = 12 [γ + , γ − ]eπ 2 {γ , (5.5.5)
with γ α = (γ c , γ c̃ ). (eiγ 5 π/4 converts to the representation where both γ A and γ A
are real.)
We now project to the OSp(1,1|2) singlet

[J AB , q 0 } = 0 → q 0 = δ(a†A aA )q , (5.5.6)

where the Kronecker δ projects down to ground states with respect to these creation
operators. It satisfies
δ(a† a)a† = aδ(a† a) = 0 . (5.5.7)
This projector can be rewritten in various forms:
du iua†A aA
δ(a†A aA ) = δ(a†α aα )a+ a− a† + a† − = e . (5.5.8)
−π 2π

We next check that this symmetry of the physical states is the usual supersym-
metry. We start with the light-cone commutation relations

{q, q̄} = 2P p/ , (5.5.9)

where P is a Weyl projector, when necessary, and, as usual, q̄ = q † η, with η the
hermitian spinor metric satisfying γ † η = ηγ. (η’s explicit form will change upon
adding dimensions because of the change in signature of the Lorentz metric.) We
then find
{q 0 , q̄0 } = δ(a† a)2P p/δ(a† a) = 2Pδ(a† a)γ a pa , (5.5.10)
where the γ A and γ A terms have been killed by the δ(a† a)’s on the left and right.
The factors other than 2γ a pa project to the physical subspace (i.e., restrict the range
of the extended spinor index to that of an ordinary Lorentz spinor). The analogous
construction for the GL(1)-type OSp(1,1|2) fails, since in that case the corresponding
projector δ(η BA γ A γ B ) = δ(η BA γ A γ B )† (whereas δ(a† a) is hermitian), and the 2 δ’s
then kill all terms in p/ except η BA γ A pB .
As a special case, we consider arbitrary massless representations of supersymme-
try. The light-cone representation of the supersymmetry generators is (cf. (5.4.20a))
q = q+ − γ + γ i pi q + , (5.5.11a)

where q + is a self-conjugate light-cone spinor:

γ −q+ = 0 , {q + , q̄ + } = 2Pγ − p+ . (5.5.11b)

Thus, q + has only half as many nonvanishing components as a Lorentz spinor, and
only half of those are independent, the other half being their conjugates. The Poincaré
algebra is then specified by
M ij = q̄γ + γ ij q + M̌ ij , (5.5.12)
where M̌ is an irreducible representation of SO(D-2), the superspin, specifying the
spin of the Clifford vacuum of q + . (Cf. (5.4.21,22). We have normalized the q̄q term
for Majorana q.)
After adding 4+4 dimensions, q + can be Lorentz-covariantly further divided using
γ ± :  
√ ∂
q + = p+ + 2γ − θ ,
∂ θ̄
∂ ∂
γ− = γ + θ = γ + = γ − θ = 0 ,
∂ θ̄ ∂ θ̄

, θ̄ = P( 12 γ − γ + )( 12 γ + γ − ) . (5.5.13)
∂ θ̄
After substitution of (5.5.11,12) into (5.5.6), we find
† √ ∂ 2
q 0 = δ(a a) p+ + √ (γ a pa + γ α pα )θ . (5.5.14)
∂ θ̄ p+
From (5.5.12) we obtain the corresponding spin operators (for Majorana θ)
∂ ∂
M ab = 12 θ̄γ ab + M̌ ab , M αβ + M α β  = θ̄a† (α aβ) + M̌ αβ + M̌ α β  ,
∂ θ̄ ∂ θ̄

M − + = − 12 θ̄ + M̌ − + , M + α = 12 θ̄γ − γ + γ α θ + M̌ + α ,
∂ θ̄
∂ 1 ∂ ∂
M αa = 12 θ̄γ α γ a
+ M̌ αa , M − α = 16 γ + γ − γ α + M̌ − α . (5.5.15)
∂ θ̄ ∂θ ∂ θ̄
Finally, we perform the unitary transformations (3.6.13) to find

q 0 = δ(a† a) + 2γ a pa θ . (5.5.16)
∂ θ̄
The δ now projects out just the OSp(1,1|2)-singlet part of θ (i.e., the usual Lorentz

q0 = + γ a pa θ 0 ,
∂ θ̄ 0
5.5. SuperBRST 113

∂ ∂ ∂
= δ(a† a) , θ0 = 2δ(a a)θ †
, , θ̄0 = Pδ(a† a) . (5.5.17)
∂ θ̄ 0 ∂ θ̄ ∂ θ̄ 0
(5.5.15) can be substituted into (3.6.14). We then find the OSp(1,1|2) generators

J +α = ixα p+ , J −+ = −ix− p+ , 
J αβ = −ix(α pβ) + M ,

1  βp + Q 

J −α = −ix− pα + −ixα 21 (pa 2 + pβ pβ ) + M α β α ; (5.5.18a)
 = θ̄a† a ∂ + M̌
M αβ (α β) αβ + M̌ α β  ,
∂ θ̄
 ∂ ∂
Q α = − i 18 + iγ − γ + γ a pa θ γ + γ − a† α − iγ − γ + γ a pa θ
∂ θ̄ ∂ θ̄

+ M̌ − α + M̌ α a pa + 12 M̌ + α pa 2 , (5.5.18b)

and J ab = −ix[a pb] + 12 θ̄γ ab ∂/∂ θ̄ + M̌ ab for the Lorentz generators. Finally, we can
remove all dependence on γ ± and γ ± by extracting the corresponding γ 0 factors
contributing to the spinor metric η:

γ 0 = −i 12 (γ + − γ − )(γ + − γ − ) = γ 0 † , δ(a† a)γ 0 = γ 0 δ(a† a) = δ(a† a) ,
∂ ∂ ∂
→ γ0 = i 12 γ − γ + , θ̄ → θ̄γ 0 = i 12 θ̄γ + γ − ,
∂ θ̄ ∂ θ̄ ∂ θ̄
∂ ∂
(γ + , γ − ) = (γ + , γ − )θ = θ̄(γ − , γ + ) = (γ − , γ + ) = 0 ,
∂ θ̄ ∂θ

, θ̄ = P( 12 γ + γ − )( 12 γ − γ + ) ; (5.5.19a)
∂ θ̄
and then convert to the harmonic oscillator basis with respect to these γ’s:
∂ † † ∂ † †
→ ea + a − , θ → 12 ea +a −
θ ,
∂ θ̄ ∂ θ̄
∂ ∂ a+ a−
θ̄ → 12 θ̄ea+ a− , → e ;
∂θ ∂θ
∂ ∂ †
a± = a± θ = θ̄a† ± = a±=0 ;
∂ θ̄ ∂θ

, θ̄ = P(a+ a− a† + a† − ) . (5.5.19b)
∂ θ̄
All a± ’s and a† ± ’s can then be eliminated, and the corresponding projection operators
(the factor multiplying P in (5.5.19b)) dropped. The only part of (5.5.18) (or the
Lorentz generators) which gets modified is

 = − 1 q̄a† d + M̌   + M̌ a p + 1 M̌   p 2
Q ,
α 2 α −α α a 2 +α a

∂ ∂
q= + γ a pa θ , d= − γ a pa θ ,
∂ θ̄ ∂ θ̄
q 0 = δ(a†α aα )q , d0 = δ(a†α aα )d . (5.5.20)
 level by level in a† ’s, we find a q at each level
If we expand the first term in Q α α

multiplying a d of the previous level. In particular, the first-level ghost q 1α multiplies
the physical d0 . This means that d0 = 0 is effectively imposed for only half of its
spinor components, since the components of q are not all independent.
An interesting characteristic of this type of BRST (as well as more conventional
BRST obtained by first-quantization) is that spinors obtain infinite towers of ghosts.
In fact, this is necessary to allow the most general possible gauges. The simplest
explicit example is BRST quantization of the action of sect. 4.5 for the Dirac spinor
quantized in a gauge where the gauge-fixed kinetic operators are all p2 instead of
p/. However, these ghosts are not all necessary for the gauge invariant theory, or for
certain types of gauges. For example, for the type of gauge invariant actions for
spinors described in sect. 4.5, the only parts of the infinite-dimensional OSp(D-1,1|2)
spinors which are not pure gauge are the usual Lorentz spinors. (E.g., the OSp(D-
1,1|2) Dirac spinor reduces to an ordinary SO(D-1,1) Dirac spinor.) For gauge-fixed,
4D N=1 supersymmetric theories, supergraphs use chains of ghost superfields which
always terminate with chiral superfields. Chiral superfields can be irreducible off-shell
representations of supersymmetry since they effectively depend on only half of the
components of θ. (An analog also exists in 6D, with or without the use of harmonic
superspace coordinates [4.12].) However, no chiral division of θ exists in 10D (θ is a
real representation of SO(9,1)), so an infinite tower of ghost superfields is necessary
for covariant background-field gauges. (For covariant non-background-field gauges,
all but the usual finite Faddeev-Popov ghosts decouple.) Thus, the infinite tower is
not just a property of the type of first-quantization used, but is an inherent property
of the second-quantized theory. However, even in background-field gauges the infinite
tower (except for the Faddeev-Popovs) contribute only at one loop to the effective
action, so their evaluation is straightforward, and the only expected problem would
be their summation.
The basic reason for the tower of θ’s is the fact that only 1/4 (or, in the massive
case, 1/2) of them appear in the gauge-invariant theory on-shell, but if θ is an irre-
ducible Lorentz representation it’s impossible to cancel 3/4 (or 1/2) of it covariantly.
We thus effectively obtain the sums
1−1+1−1+··· = 2 , (5.5.21a)
5.5. SuperBRST 115

1−2+3−4+··· = 4 . (5.5.21b)
(The latter series is the “square” of the former.) The positive contributions rep-
resent the physical spinor (or θ) and fermionic ghosts at even levels, the negative
contributions represent bosonic ghosts at odd levels (contributing in loops with the
1 1
opposite sign), and the 2 or 4 represents the desired contribution (as obtained directly
in light-cone gauges). Adding consecutive terms in the sum gives a nonconvergent
(but nondivergent in case (5.5.21a)) result which oscillates about the desired result.
However, there are unambiguous ways to regularize these sums. For example, if
we represent the levels in terms of harmonic oscillators (one creation operator for
(5.5.21a), and the 2 a† α ’s for (5.5.21b)), these sums can be represented as integrals
over coherent states (see (9.1.12)). For (5.5.21a), we have:

d2 z −|z|2  
str(1) = tr (−1)N = e z (−1)a a z
 2  2
d z −|z|2 d z −2|z|2 1  d2 z −|z|2
= e z| − z
= e =2 e
π π π
= 12 , (5.5.22)

where str is the supertrace; for (5.5.21b), the supertrace over the direct product
corresponding to 2 sets of oscillators factors into the square of (5.5.22). (The corre-
sponding partition function is str(xN ) = 1/(1 + x) = 1 − x + x2 − · · ·, and for 2 sets
of oscillators 1/(1 + x)2 = 1 − 2x + 3x2 − · · ·.)
An interesting consequence of (5.5.21) is the preservation of the identity

D  = 2(D−k)/2 ; D  = str S (1) , D = str V (1) ; (5.5.23)

upon adding (2, 2|4) dimensions, where D  and D are the “superdimensions” of a
spinor and vector, defined in terms of supertraces of the identity for that representa-
tion, and k is an integer which depends on whether the dimension is even or odd and
whether the spinor is Weyl and/or Majorana (see sect. 5.3). k is unchanged by adding
(2, 2|4) dimensions, D changes by addition of 4 − 4 = 0, and, because of (5.5.21), D 
changes by a factor of 22 · ( 12 )2 = 1. This identity is important for super-Yang-Mills
and superstrings.
Before considering the action for arbitrary supersymmetric theories, we’ll first
study the equations of motion, since the naive kinetic operators may require extra
factors to write a suitable lagrangian. Within the OSp(1,1|2) formalism, the gauge-
fixed field equations are (cf. (4.4.19))

(p2 + M 2 )φ = 0 (5.5.24a)

when subject to the (Landau) gauge conditions [2.3]
Q  φ=0 . (5.5.24b)
α αβ

Applying these gauge conditions to the gauge transformations, we find the residual
gauge invariance

δφ = −i 12 Q ,  2 Λα = M
− 32 (p2 + M 2 ) + Q  Λ + C Λ = 0 . (5.5.24c)
α αβ γ γ(α β)

(In the IGL(1) formalism, sect. 4.2, the corresponding equations involve just the Q
 + and M
component of Qα and the M  3 components of M
αβ , but are equivalent, since 
+ = M
M 3 = 0 → M
 − = 0, and Q
+ = M
− = 0 → Q
 − = 0.)

For simplicity, we consider the massless case, and M̌ ij = 0. We can then choose
the reference frame where pa = δ a + p+ , and solve these equations in light-cone nota-
tion. (The +’s and −’s now refer to the usual Lorentz components; the unphysical x− ,
p+ , and γ ± have already been eliminated.) The gauge conditions (5.5.24b) eliminate
auxiliary degrees of freedom (as ∂ · A = p+ A− = 0 eliminates A− in Yang-Mills),
and (5.5.24c) eliminates remaining gauge degrees of freedom (as A+ in light-cone-
gauge Yang-Mills). We divide the spinors d, q, ∂/∂θ, and θ into halves using γ ± , and
then further divide those into complex conjugate halves as creation and annihilation
operators, as in (5.4.27,29):

d → γ + d, γ − d → da , d¯a , ∂ a , ∂¯a ,

q → γ + q, γ − q → q a , q̄a , ∂ a , ∂¯a , (5.5.25)
where the “−” parts of d and q are both just partial derivatives because the mo-
mentum dependence drops out in this frame, and d, d¯ and q, q̄ have graded harmonic
oscillator commutators (up to factors of p+ ). (5.5.20) then becomes
 ∼ q̄ a a† ∂ + q a† ∂¯a + ∂¯a a† d + ∂ a† d¯a
Q . (5.5.26)
α α a a α α a a α

 consists of terms of the form AB, either A or B can be chosen as the con-
Since Qα

straint in (5.5.24b), and the other will generate gauge transformations in (5.5.24c).
We can thus choose either ∂ a , or d¯a and q̄a , and similarly for the complex conjugates,
except for the Sp(2) singlets, where the choice is between ∂ a and just d¯a (and simi-
larly for the complex conjugates). However, choosing both d and d¯ (or both q and q̄)
for constraints causes the field to vanish, and choosing them both for gauge genera-
tors allows the field to be completely gauged away. As a result, the only consistent
constraints and gauge transformations are

d¯a φ = (aα q̄ a )φ = ∂¯a φ = 0 , δφ = ∂ a λ̄a , (5.5.27a)
5.5. SuperBRST 117

subject to the restriction that the residual gauge transformations preserve the gauge
choice (explicitly, (5.5.24c), although it’s more convenient to re-solve for the residual
invariance in light-cone notation). (There is also a complex conjugate term in φ if it
satisfies a reality condition. For each value of the index a, the choice of which oscillator
is creation and which is annihilation is arbitrary, and corresponding components of d
and d¯ or q and q̄ can be switched by changing gauges.) Choosing the gauge

θ−a φ = 0 , (5.5.27b)

for the residual gauge transformation generated by ∂ a = ∂/∂θ−a , the field becomes

φ(θ+ , θ̄+ , θ− , θ̄− ) = δ(aα θ+ )δ(θ− )ϕ(θ0 + , θ̄0 + ) , d¯0 ϕ = 0 . (5.5.28)

ϕ is the usual chiral light-cone superfield (as in sect. 5.4), a function of only 1/4 of
the usual Lorentz spinor θ0 . This agrees with the general result of equivalence to the
light cone for U(1)-type 4+4-extended BRST given at the end of sect. 4.5.

Since the physical states again appear in the middle of the θ expansion (including
ghost θ’s), we can again use (4.1.1) as the action: In the light-cone gauge, from
(5.5.28), integrating over the δ-functions,

S= dx dθ0 + dθ̄0 + ϕ̄2 ϕ , (5.5.29)

which is the standard light-cone superspace action [5.25]. As usual for the expansion
of superfields into light-cone superfields, the physical light-cone superfield appears in
the middle of the non-light-cone-θ expansion of the gauge superfield, with auxiliary
light-cone superfields appearing at higher orders and pure gauge ones at lower orders.
Because some of the ghost θ’s are commuting, we therefore expect an infinite num-
ber of auxiliary fields in the gauge-covariant action, as in the harmonic superspace
formalism [4.12]. This may be necessary in general, because this treatment includes
self-dual multiplets, such as 10D super-Yang-Mills. (This multiplet is superspin 0,
and thus does not require the superspin M̌ ij to be self-dual, so it can be treated in the
OSp(1,1|2) formalism. However, an additional self-conjugacy condition on the light-
cone superfield is required, (5.4.31b), and a covariant OSp(1,1|2) statement of this
condition would be necessary.) However, in some cases (such as 4D N=1 supersym-
metry) it should be possible to truncate out all but a finite number of these auxiliary
superfields. This would require an (infinite) extension of the group OSp(1,1|2) (per-
haps involving part of the unphysical supersymmetries aα q), in the same way that
extending IGL(1) to OSp(1,1|2) eliminates Nakanishi-Lautrup auxiliary fields.

The unusual form of the OSp(1, 1|2) operators for supersymmetric particles may
require new mechanics actions for them. It may be possible to derive these actions
by inverting the quantization procedure, first using the BRST algebra to derive the
hamiltonian and then finding the gauge-invariant classical mechanics lagrangian.


(1) Derive (5.1.2) from (5.1.1) and (3.1.11).
(2) Show that, under the usual gauge transformation A → A + ∂λ, exp[−iq
dτ x ·
A(x)] transforms with a factor exp{−iq[λ(x(τ f )) − λ(x(τ i ))]}. (In a Feynman
path integral, this corresponds to a gauge transformation of the ends of the prop-
(3) Fourier transform (5.1.13), using (5.1.14). Explicitly evaluate the proper-time
integral in the massless case to find the coordinate-space Green function satisfying
2 G(x, x ) = δ D (x − x ) for arbitrary D > 2. Do D = 2 by differentiating with
respect to x2 , then doing the proper-time integral, and finally integrating back
with respect to x2 . (There is an infinite constant of integration which must be
renormalized.) For comparison, do D = 2 by taking the limit from D > 2.
(4) Use the method described in (5.1.13,14) to evaluate the 1-loop propagator correc-
tion in φ3 theory. Compare the corresponding calculation with the covariantized
light-cone method of sect. 2.6. (See exercise (7) of that chapter.)
(5) Derive (5.2.1), and find J 3 . Derive (5.2.2) and the rest of the OSp(1,1|2) algebra.
Show these results agree with those of sect. 3.4.
(6) Quantize (5.3.10) in the 3 supersymmetric gauges described in that section, and
find the corresponding IGL(1) (and OSp(1,1|2), when possible) algebras in each
case, using the methods of sects. 3.2-3. Note that the methods of sect. 3.3 require
some generalization, since commuting antighosts can be conjugate to the corre-
sponding ghosts and still preserve Sp(2): [C α , C β ] ∼ C αβ . Show equivalence to
the appropriate algebras of sects. 3.4-5.
(7) Derive the tables in sect. 5.3. (Review the group theory of SO(N) spinors, if
necessary.) Use the tables to derive the groups, equivalent to SO(D + ,D − ) for
D ≤ 6, for which these spinors are the defining representation.
(8) Express the real-spinor γ-matrices of sect. 5.3 in terms of the σ-matrices there for
arbitrary D + and D − . Use the Majorana representation where the spinor is not
Exercises 119

necessarily explicitly real, but equivalent to a real one, such that: (1) for complex
representations, the bottom half of the spinor is the complex conjugate of the top
half (each being irreducible); (2) for pseudoreal representations, the bottom half
is the complex conjugate again but with the index converted with a metric to
make it explicitly the same representation as the top; (3) for real representations,
the spinor is just the real, irreducible one. Find the matrices representing the
internal symmetry (U(1) for complex and SU(2) for pseudoreal).

(9) Check the Jacobi identities for the covariant derivatives whose algebra is given
in (5.4.8). Check closure of the algebra (5.4.10) in D = 10, including the extra
generator described in the text.

(10) Derive (5.4.13).

(11) Write the explicit action and transformation laws for (5.4.16).

(12) Write the explicit equations of motion (5.4.12), modified by (5.4.23), for a scalar
superfield for N=1 supersymmetry in D=4. Show that this gives the usual co-
variant constraints and field equations (up to constants of integration) for the
chiral scalar superfield (scalar multiplet). Do the same for a spinor superfield,
and obtain the equations for the vector-multiplet field strength.

(13) Derive the explicit form of the twistor fields for D=3,4,6. Find an explicit ex-
pression for the supersymmetrized Pauli-Lubansky vector in D=4 in terms of
supertwistors, and show that it automatically gives an explicit expression for the
superhelicity H of (5.4.23) as an operator in supertwistor space. Show the su-
pertwistor Pauli-Lubansky vector automatically vanishes in D=3, and derive an
expression in D=6.


6.1. Gauge covariant

In this chapter we’ll consider the mechanics action for the string and its gauge
fixing, as a direct generalization of the treatment of the particle in the previous
The first-order action for string mechanics is obtained by generalizing the 1-
dimensional particle mechanics world-line of (5.1.1) to a 2-dimensional world sheet

1 d2 σ
S=  (∂ m X) · P m + gmn 12 P m · P n , (6.1.1)
α 2π
where X(σ m ) is the position in the higher-dimensional space in which the world
sheet is imbedded of the point whose location in the world sheet itself is given by
σ m = (σ 0 , σ 1 ) = (τ, σ), d2 σ = dσ 0 dσ 1 = dτ dσ, ∂ m = (∂ 0 , ∂ 1 ) = (∂/∂τ, ∂/∂σ), and
gmn = (−g)−1/2 g mn is the unit-determinant part of the 2D metric. (Actually, it has
determinant −1.) 1/2πα is both the string tension and the rest-mass per unit length.
(Their ratio, the square of the velocity of wave propagation in the string, is unity in
units of the speed of light: The string is relativistic.) This action is invariant under
2D general coordinate transformations (generalizing (5.1.4); see sect. 4.1):

δX =
m ∂ m X ,

δP m = ∂ n (
n P m ) − P n ∂ n
m ,

δgmn = ∂ p (
p gmn ) − gp(m ∂ p
n) . (6.1.2)

Other forms of this action which result from eliminating various combinations of
the auxiliary fields P m and gmn are
S = 

d2 σ .

X ·P 0

1 1 02
+ X 2 ) −
g01 0
P · X

2 (P (6.1.3a)
α 2π g11 g11
1  d2 σ
X · P 0 − λ+ 41 (P 0 + X  )2 − λ− 41 (P 0 − X  )2

α 2π
6.1. Gauge covariant 121

1 d2 σ mn 1
=−  g 2 (∂ m X) · (∂ n X) (6.1.3c)
α 2π
1  d2 σ √
=  (∂ m X) · P m + −det P m · P n (6.1.3d)
α 2π
1 dσ 1
=−  −det (∂ m X) · (∂ n X) = −  −(dX a ∧ dX b )2 ,
α 2π α

1 1  dσ
where α
P0 is the momentum (σ-)density (the momentum p = α 2π
P 0 ),  = ∂/∂σ,
and to obtain (6.1.3d) we have used the determinant of the g equations of motion

P m · P n − 12 gmn gpq P p · P q = 0 . (6.1.4)

As a consequence of this equation and the equation of motion for P , the 2D metric is
proportional to the “induced” metric ∂ m X ·∂ n X (as appears in (6.1.3e)), which results
from measuring distances in the usual Minkowski way in the D-dimensional space in
which the 2D surface is imbedded (using dX = dσ m ∂ m X). (The equations of motion
don’t determine the proportionality factor, since only the unit-determinant part of
the metric appears in the action.) In analogy to the particle, (6.1.4) also represents
the generators of 2D general-coordinate transformations. (6.1.3a) is the hamiltonian
form, (6.1.3b) is a rewriting of the hamiltonian form to resemble the example (3.1.14)
(but with the indefinite-metric sum of squares of both left- and right-handed modes
constrained), (6.1.3c) is the second-order form, and (6.1.3e) is the area swept out by
the world sheet [6.2]. If the theory is derived from the form (6.1.3b), there are 2 sets
of transformation laws of the form (3.1.15) (with appropriate sign differences), and
(6.1.2,3c) can then be obtained as (3.1.16,17).

For the open string, the X equations of motion also imply certain boundary
conditions in σ. (By definition, the closed string has no boundary in σ.) Varying the
(∂X) · P term and integrating by parts to pull out the δX factor, besides the equation
of motion term ∂P we also get a surface term nm P m , where nm is the normal to the
boundary. If we assume 2D coordinates such that the position of the σ boundaries
are constant in τ , then we have the boundary condition

P1 = 0 . (6.1.5)

It’s convenient to define the quantities
P (±) = √ (P 0 ± X  ) (6.1.6a)

because the hamiltonian constraints appearing in (6.1.3b) (equivalent to (6.1.4)) can
be expressed very simply in terms of them as

P (±)2 = 0 , (6.1.6b)

and because they have simple Poisson brackets with each other. For the open string,
it’s further useful to extend σ: If we choose coordinates such that σ = 0 for all τ at
one end of the string, and such that (6.1.5) implies X  = 0 at that end, then we can
X(σ) = X(−σ) , P (σ) = √ 
(P 0 + X  ) = P (±) (±σ) f or ± σ > 0 , (6.1.7a)

so the constraint (6.1.6b) simplifies to

P 2 = 0 . (6.1.7b)

6.2. Conformal gauge

The conformal gauge is given by the gauge conditions (on (6.1.1,3ac))

gmn = η mn , (6.2.1)

where η is the 2D flat (Minkowski) space metric. (Since g is unit-determinant, it has
only 2 independent components, so the 2 gauge parameters of (6.1.2) are sufficient
to determine it completely.) As for the particle, this gauge can’t be obtained every-
where, so it’s imposed everywhere except the boundary in τ . Then variation of g at
initial or final τ implies (6.1.4) there, and the remaining field equations then imply it
everywhere. In this gauge those equations are

P m = −∂ m X , ∂mP m = 0 → ∂2X = 0 . (6.2.2)

It is now easy to see that the endpoints of the string travel at the speed of light. (With

. .
slight generalization, this can be shown in arbitrary gauges.) From (6.1.4,5) and
(6.2.1,2), we find that dX = dτ X, and thus dX · dX = dτ 2 X 2 = dτ 2 (X 2 + X 2 ) = 0.
(6.2.2) is most easily solved by the use of 2D light-cone coordinates
± 1 0 1
σ = √ (σ 1 ∓ σ 0 ) → η mn = = η mn , (6.2.3)
2 1 0
where 2D indices now take the values ±. We then have

∂ + ∂ − X = 0 → X = 12 [X̂ (+) (τ + σ) + X̂ (−) (τ − σ)] . (6.2.4a)
6.2. Conformal gauge 123

For the open string, the boundary condition at one boundary, chosen to be σ = 0, is
.̂ (+)

(∂ + + ∂ − )X = 0 → X (τ ) = X (−) (τ ) → X̂ (+) (τ ) = X̂ (−) (τ ) , (6.2.4b)

without loss of generality, since the constant parts of X̂ (±) appear in X only as their
sum. Thus, the modes of the open string correspond to the modes of one handedness
of the closed string. The boundary condition at the other boundary of the open
string, taken as σ = π, and the “boundary” condition of the closed string, which is
simply that the “ends” at σ = ±π are the same point (and thus the closed string X
is periodic in σ with period 2π, or equivalently X and X  have the same values at
σ = π as at σ = −π) both take the form
.̂ (±)

X (τ + 2π) = X (±) (τ ) → X̂ (±) (τ + 2π) = X̂ (±) (τ ) + 4παp(±) , (6.2.4c)

p(+) = p(−) . (6.2.4d)

The constraints (6.1.4) also simplify:

P ±2 (τ, σ) = 12 X̂ (±)2 (τ ± σ) = 0 . (6.2.5)

These constraints will be used to build the BRST algebra in chapt. 8.

The fact that the modes of the open string correspond to half the modes of
the closed string (except that both have 1 zero-mode) means that the open string
can be formulated as a closed string with modes of one handedness (clockwise or
counterclockwise). This is accomplished by adding to the action (6.1.1) for the closed
string the term

d2 σ 1
S1 = ua ub λmn 12 (gmp −
mp ) 12 (gnq −
nq )(∂ p X a )(∂ q X b ) , (6.2.6)
2π 2
where u is a constant, timelike or lightlike (but not spacelike) vector (u2 ≤ 0), and

+− = −1. λ is a lagrange multiplier which constrains (u · ∂ − X)2 = 0, and thus
u · ∂ − X = 0, in the gauge (6.2.1). Together with (6.2.5), this implies the Lorentz
covariant constraint ∂ − X a = 0, so X depends only on τ − σ, as in (3.1.14-17). Thus,
the formulation using (6.2.6) is Lorentz covariant even though S 1 is not manifestly
so (because of the constant vector u). We can then identify the new X with the X̂ of
(6.2.4). Since λ itself appears multiplied by ∂ − X in the equations of motion, it thus
drops out, implying that it’s a gauge degree of freedom which, like g, can be gauged
away except at infinity.

Similar methods can be applied to the one-handed modes of the heterotic string
[1.13]. (Then in (6.2.6) only spacelike X’s appear, so instead of ua ub any positive-
definite metric can be used, effectively summing over the one-handed X’s.) Various
properties of the actions (3.1.17, 6.2.6) have been discussed in the literature [6.3], par-
ticularly in relation to anomalies in the gauge symmetry of the lagrange multipliers
upon naive lagrangian quantization. One simple way to avoid these anomalies while
keeping a manifestly covariant 2D lagrangian is to add scalars φ with the squares of
both ∂ − φ and ∂ + φ appearing in lagrange-multiplier terms [6.4]. Alternatively (or ad-
ditionally), one can add Weyl-Majorana 2D spinors (i.e., real, 1-component, 1-handed
spinors) whose nonvanishing energy-momentum tensor component couples to the ap-
propriate lagrange multiplier. (E.g., a spinor with kinetic term ψ∂ + ψ appears also
in the term λ[(∂ − φ)2 + ψ∂ − ψ].) These nonpropagating fields appear together with
scalars with only one or the other handedness or neither constrained, and uncon-
strained fermions which are Weyl and/or Majorana or neither. There are (at most)
2 lagrange multipliers, one for each handedness.
In the conformal gauge there is still a residue of the gauge invariance, which origi-
nally included not only 2D general coordinate transformations but also local rescalings
of the 2D metric (since only its unit-determinant part appeared in the action). By
definition, the subset of these transformations which leave (6.2.1) invariant is the con-
formal group. Unlike in higher dimensions, the 2D conformal group has an infinite
number of generators. It can easily be shown that these transformations consist of
the coordinate transformations (restricted by appropriate boundary conditions)

σ ± = ζ ± (σ ± ) , (6.2.7a)

with ±’s not mixing (corresponding to 2 1D general coordinate transformations), since
these coordinate transformations have an effect on the metric which can be canceled
by a local scaling:
dσ 2 = 2dσ + dσ − = ζ + ζ − 2dσ + dσ − . (6.2.7b)
On shell, these transformations are sufficient to gauge away one Lorentz component
of X, another being killed by the constraint (6.2.5). These 2 Lorentz components can
be eliminated more directly by originally choosing stronger gauge conditions, as in
the light-cone gauge.
The conformal gauge is a temporal gauge, since it is equivalent to setting the
time components of the gauge field to constants: gm0 = η m0 . When generalized to
D > 2, it is the choice of Gaussian normal coordinates. We can instead choose a
6.3. Light cone 125

Lorentz gauge, ∂ m gmn = 0. This is the De Donder gauge, or harmonic coordinates,
which is standardly used in D > 2. We’ll discuss this gauge in more detail in sect.

6.3. Light cone
In a light-cone formalism [6.5] not only are more gauge degrees of freedom elim-
inated than in covariant gauges, but also more (Lorentz) auxiliary fields. We do
the latter first by varying the action (6.1.1) with respect to all fields carrying a “−”
Lorentz index (X − , P m − ):
→ ∂ m P m+ = 0 ; (6.3.1a)
δX −
→ gmn = (Ar B r )−1 (
pm Ap
qn Aq − B m B n ) ,
δP m −
Am = P m + , Bm = ∂mX + . (6.3.1b)

We next eliminate all fields with a “+” index by gauge conditions:

τ: X + = kτ

σ: P 0+ = k , (6.3.2a)

where k is an arbitrary constant. (The same procedure is applied in light-cone Yang-
Mills, where A− is eliminated as an auxiliary field and A+ as a gauge degree of
freedom: see sect. 2.1.) The latter condition determines σ to be proportional to the
amount of +-momentum between σ = 0 and the point at that value of σ (so the string

length is proportional to dσP 0 + , which is a constant, since ∂ m P m + = 0). Thus,
σ is determined up to a function of τ (corresponding to the choice of where σ = 0).
However, P 1 + is also determined up to a function of τ (since now ∂ 1 P 1 + = 0), so σ
is completely determined, up to global translations σ → σ + constant, by the further
P 1+ = 0 . (6.3.2b)
(6.3.2) implies (6.2.1). For the open string, by the converse of the argument leading
to the boundary condition (6.1.5), this determines the values of σ at the boundaries
up to constants, so the remaining global invariance is used to choose σ = 0 at one
boundary. For the closed string, the global invariance remains, and is customarily
dealt with in the quantum theory by imposing a constraint of invariance under this
transformation on the field or first-quantized wave function.

1  dσ 0
The length of the string is then given by integrating (6.3.2a): p+ = α 2π
P + =

(1/2πα )k · length. The two most convenient choices are

k=1 → length = 2πα p+

length = π (2π) → k = 2α p+ (α p+ ) f or open (closed) . (6.3.3)

For the free string the latter choice is more convenient for the purpose of mode
expansions. In the case of interactions k must be constant even through interactions,
and therefore can’t be identified with the value of p+ of each string, so the former
choice is made. Note that for p+ < 0 the string then has negative length. It is then
interpreted as an antistring (or outgoing string, as opposed to incoming string). The
use of negative lengths is particularly useful for interactions, since then the vertices
are (cyclically) symmetric in all strings: e.g., a string of length 1 breaking into 2
1 1
strings of length 2 is equivalent to a string of length 2 breaking into strings of length
1 and − 12 .

The action now becomes

− p+ + 

(∂ m X i )P m i + η mn 12 P m i P n i

, (6.3.4a)
α 2π

or, in hamiltonian form,



dσ .
X i P 0 i − 12 (P 0 i 2 + X  i 2 )

− p+ +  . (6.3.4b)
α 2π

X i is found as in (6.2.4), but X + is given by (6.3.2), and X − is given by varying the
original action with respect to the auxiliary fields gmn and P m + , conjugate to those
varied in (6.3.1):

→ P m − = −P m i P 0 i − 12 δ 0 m η np P ni P pi

δ dσ 1
→ X − = x− − P − + constant , (6.3.5)
δP + 2π

1  dσ
. 0 .
where the constant is chosen to cancel the integral when integrated over σ, so that
α 2π −P + = x− p+ in (6.3.4). (This term has been restored in (6.3.4) in order
to avoid using equations of motion to eliminate any coordinates whose equations of
motion involve time derivatives, and are thus not auxiliary. In (6.3.1a) all but the
zero-mode part of the X − equation can be used to solve for all but a σ-independent
part of P 1 + , without inverting time derivatives.)
Exercises 127

After the continuous 2D symmetries have been eliminated by coordinate choices,
certain discrete symmetries remain: σ and τ reversal. As in the particle case, τ re-
versal corresponds to a form of charge conjugation. However, the open string has
a group theory factor associated with each end which is the complex-conjugate rep-
resentation of that at the other end (so for unitary representations they can cancel
for splitting or joining strings), so for charge conjugation the 2 ends should switch,
which requires σ reversal. Furthermore, the closed string has clockwise and coun-
terclockwise modes which are distinguishable (especially for the heterotic string), so
again we require σ reversal to accompany τ reversal to keep σ ± from mixing. We
therefore define charge conjugation to be the simultaneous reversal of τ and σ (or
σ → 2παp+ − σ to preserve the positions of the boundaries of the open string). On
the other hand, some strings are nonoriented (as opposed to the oriented ones above)
in that solutions with σ reversed are not distinguished (corresponding to open strings
with real representations for the group theory factors, or closed strings with clock-
wise modes not separated from counterclockwise). For such strings we also need to
define a σ reversal which, because of its action on the 2D surface, is called a “twist”.
In the quantum theory these invariances are imposed as constraints on the fields or
first-quantized wave functions (see chapt. 10).


(1) Derive (6.1.3) from (6.1.1). Derive (6.1.5).

(2) Derive (6.3.1).


7.1. Bosonic

In this section we will quantize the light-cone gauge bosonic string described in
sect. 6.3 and derive the Poincaré algebra, which is a special case of that described in
sect. 2.3.

The quantum mechanics of the free bosonic string is described in the light-cone
formalism [6.5] in terms of the independent coordinates X i (σ) and x− , and their
canonical conjugates P 0 i (σ) and p+ , with “time” (x+ = 2α p+ τ ) dependence given
by the hamiltonian (see (6.3.4b))

1 dσ 1 0 2
H= (P i + X i 2 ) . (7.1.1)
α 2π 2
1 0 δ δ
P i (σ) = i , , X j (σ 2 ) = δ ij 2πδ(σ 2 − σ 1 ) . (7.1.2)
α δX i (σ) δX i (σ 1 )

(Note our unconventional normalization for the functional derivative.)

For the open string, it’s convenient to extend σ from [0, π] to [−π, π] as in (6.1.7)
by defining
1 δ 1
X(−σ) = X(σ) , P (σ) = √  α i + X  = √  P ± (±σ) f or ± σ > 0 ,
2α δX α
so that P is periodic (and ∼ P + or P − , as in (6.2.5), for σ > or < 0), or to express
the open string as a closed string with modes which propagate only clockwise (or only
counterclockwise) in terms of X̂ of sect. 6.2:
P (σ) = √ X̂  (σ) , (7.1.3b)
7.1. Bosonic 129

which results in the same P (same boundary conditions and commutation relations).
The latter interpretation will prove useful for graph calculations. (However, the form
of the interactions will still look different from a true closed string.) Either way, the
new boundary conditions (periodicity) allow (Fourier) expansion of all operators in
terms of exponentials instead of cosines or sines. Furthermore, P contains all of X

and δ/δX (except x, which is conjugate to p ∼ dσ P ; i.e., all the translationally
invariant part). In particular,
dσ 1  2
H= Pi . (7.1.4)
−π 2π 2
The commutation relations are

[P i (σ 1 ), P j (σ 2 )] = 2πiδ  (σ 2 − σ 1 )δ ij . (7.1.5)

For the closed string, we define 2 P ’s by
1 δ 1
P (±) (σ) = √  α i ± X  (±σ) = √  X̂ (±) (σ) . (7.1.6)
2α δX(±σ) 2α
Then operators which are expressed in terms of integrals over σ also include sums
over ±: e.g., H = H (+) + H (−) , with H (±) given in terms of P (±) by (7.1.4).
In order to compare with particles, we’ll need to expand all operators in Fourier
modes. In practical calculations, functional techniques are easier, and mode expan-
sions should be used only as the final step: external line factors in graphs, or expansion
of the effective action. (Similar remarks apply in general field theories to expansion of
superfields in components and expansion of fields about vacuum expectation values.)
For P for the open string (suppressing Lorentz indices),

P (σ) = αn e−inσ ,
√ √
α0 = 2αp , αn = (α−n )† = −i nan † ,

[p, x] = i , [αm , αn ] = nδ m+n,0 , [am , an † ] = δ mn . (7.1.7a)

We also have
√ ∞
X̂(σ) = X̂ † (σ) = X̂ *(−σ) = (x + 2α pσ) + 2α √ (an † e−inσ + an einσ ) ,
1 n
X(σ) = 1
2 X̂(σ) + X̂(−σ) , P 0 (σ) = α i = 12 X̂  (σ) + X̂  (−σ) .

(The and * have the usual matrix interpretation if the operators are considered

as matrices acting on the Hilbert space: The is the usual operatorial hermitian
conjugate, whereas the * is the usual complex conjugate as for functions, which
changes the sign of momenta. Combined they give the operatorial transpose, which
corresponds to integration by parts, and thus also changes the sign of momenta, which
are derivatives.) H is now defined with normal ordering:

H = 2 : αn · α−n : + constant

= α  pi 2 + nan † · an + constant = α pi 2 + N + constant . (7.1.8a)

In analogy to ordinary field theory, i∂/∂τ + H = α p2 + N + constant.
The constant in H has been introduced as a finite renormalization after the infi-
nite renormalization done by the normal ordering: As in ordinary field theory, wher-
ever infinite renormalization is necessary to remove infinities, finite renormalization
should also be considered to allow for the ambiguities in renormalization prescrip-
tions. However, also as in ordinary field theories, the renormalization is required
to respect all symmetries of the classical theory possible (otherwise the symmetry is
anomalous, i.e., not a symmetry of the quantum theory). In the case of any light-cone
theory, the one symmetry which is never manifest (i.e., an automatic consequence of
the notation) is Lorentz invariance. (This sacrifice was made in order that unitarity
would be manifest by choosing a ghost-free gauge with only physical, propagating
degrees of freedom.) Thus, in order to prove Lorentz invariance isn’t violated, the
1  dσ 0
commutators of the Lorentz generators J ab = α 2π
X [a P b] must be checked. All
are trivial except [J −i , J −j ] = 0 because X − and P 0 − are quadratic in X i and P 0 i by
(6.3.5). The proof is left as an (important) exercise for the reader that the desired
result is obtained only if D = 26 and the renormalization constant in H and P 0 −
 dσ 0
(H = − α1 2π
P −) is given by

H = α pi 2 + N − 1 ≡ α (pi 2 + M 2 ) . (7.1.8b)

The constant in H also follows from noting that the first excited level (a1i † |0
) is
just a transverse vector, which must be massless. The mass spectrum of the open
string, given by the operator M 2 , is then a harmonic oscillator spectrum: The possible
values of the mass squared are −1, 0, 1, 2, ... times 1/α , with spins at each mass
level running as high as N = α M 2 + 1. (The highest-spin state is created by the
symmetric traceless part of N a1 † ’s.)
7.1. Bosonic 131

Group theory indices are associated with the ends of the open string, so the
string acts like a matrix in that space. If the group is orthogonal or symplectic,
the usual (anti)symmetry of the adjoint representation (as a matrix acting on the
vector representation) is imposed not just by switching the indices, but by flipping
the whole string (switching the indices and σ ↔ π − σ). As a result, N odd gives
matrices with the symmetry of the adjoint representation (by including an extra “−”
sign in the condition to put the massless vector in that representation), while N even
gives matrices with the opposite symmetry. (A particular curiosity is the “SO(1)”
open string, which has only N even, and no massless particles.) Such strings, because
of their symmetry, are thus “nonoriented”. On the other hand, if the group is just
unitary, there is no symmetry condition, and the string is “oriented”. (The ends can
be labeled with arrows pointing in opposite directions, since the hermitian conjugate
state can be thought of as the corresponding “antistring”.)

The closed string is treated analogously. The mode expansions of the coordinates
and operators are given in terms of 2 sets of hatted operators X̂ (±) or P (±) expanded
over α(±) n :

X(σ) = 1
2 X̂ (+) (σ) + X̂ (−) (−σ) , P 0 (σ) = 1
2 X̂ (+) (σ) + X̂ (−) (−σ) ;

X̂ (±) (σ) = X̂ (±)† (σ) = X̂ (±) *(−σ) → α(±) n = α(±) −n † = −α(±) n * . (7.1.9)

However, because of the periodicity condition (6.2.4c), and since the zero-modes (from
(7.1.7)) appear only as their sum, they are not independent:

x(±) = x , p(±) = 12 p . (7.1.10)

Operators which have been integrated over σ (as in (7.1.4)) can be expressed as sums
over two sets (±) of open-string operators: e.g.,

H = H (+) + H (−) = 12 α pi 2 + N − 2 , N = N (+) + N (−) . (7.1.11)

Remembering the constraint under global σ translations

d dσ  δ
i ≡ X ·i = ∆N ≡ N (+) − N (−) = 0 , (7.1.12)
dσ 2π δX
the spectrum is then given by the direct product of 2 open strings, but with the states
constrained so that the 2 factors (from the 2 open strings) give equal contributions
to the mass. The masses squared are thus −4, 0, 4, 8, ... times 1/α , and the highest
spin at any mass level is N = 12 α M 2 + 2 (with the corresponding state given by the

symmetric traceless product of N a1 † ’s, half of which are from one open-string set
and half from the other). Compared with α M 2 + 1 for the open string, this means
that the “leading Regge trajectory” N(M 2 ) (see sec. 9.1) has half the slope and twice
the intercept for the closed string as for the open string. If we apply the additional
constraint of symmetry of the state under interchange of the 2 sets of string operators,
the state is symmetric under interchange of σ ↔ −σ, and is therefore “nonoriented”;
otherwise, the string is “oriented”, the clockwise and counterclockwise modes being
distinguishable (so the string can carry an arrow to distinguish it from a string that’s
flipped over).

From now on we choose units

α = 1
2 . (7.1.13)

In the case of the open bosonic string, the free light-cone Poincaré generators can
be obtained from the covariant expressions (the obvious generalization of the particle
expressions, because of (7.1.2), since X a and P 0 a are defined to Lorentz transform as
vectors and X to translate by a constant),
dσ dσ 0
J ab = −i X [a (σ)P 0 b] (σ) , pa = P a (σ) , (7.1.14a)
−π 2π 2π
by substituting the gauge condition (6.3.2) and free field equations (6.1.7) (using
P i 2 − 2
X̂ + (σ) = p+ σ , P 2 − 2 = 0 → P − = −

→ J ij = −ix[i pj] + a† n[i anj] ,

J +i = ixi p+ , J −+ = −ix− p+ ,

J −i = −i(x− pi − xi p− ) + a† n[− ani] ,
1 2
1 i
p− = − pi + 2 na† n i ani − 1 , an− = − (p ani + ∆n ) ,
2p+ p+
 n−1 ∞

1 †
∆n = i √ 2 m(n − m)am an−m,i −
m(n + m)a m
an+m,i .
n m=1 m=1
(We have added the normal-ordering constant of (7.1.8) to the constraint (6.1.7).)
As for arbitrary light-cone representations, the Poincaré generators can be expressed
completely in terms of the independent coordinates xi and x− , the corresponding
momenta pi and p+ , the mass operator M, and the generators of a spin SO(D−1),
7.1. Bosonic 133

M ij and M im . For the open bosonic string, we have the usual oscillator representation
of the SO(D−2) spin generators
M ij = −i X̂ i P j = a† n[i anj] , (7.1.15a)

where means the zero-modes are dropped. The mass operator M and the remaining
SO(D−1) operators M im are given by

M = −2p+ P − = (P i 2 − 2) = 2 na† n i ani − 1 = 2(N − 1) ,

M im M = ip+ X̂ i P − = a† ni ∆n − ∆† n ani . (7.1.15b)

The usual light-cone formalism for the closed string is not a true light-cone for-
malism, in the sense that not all constraints have been solved explicitly by eliminating
variables: The one constraint that remains is that the contribution to the “energy”
p− from the clockwise modes is equal to that from the counterclockwise ones. As a
result, the naive Poincaré algebra does not close [4.10]: Using the expressions

J −i = −i(x− pi − xi p− ) + a†(±) n[− a(±) ni] ,

1 2

p− = − pi + 4 N (+) + N (−) − 2 , (7.1.16a)
we find
[J −i , J −j ] = − ∆N∆J ij , (7.1.16b)
p+ 2
where ∆J ij is the difference between the (+) and (−) parts of J ij .

We instead define 2 sets of open-string light-cone Poincaré generators J (±) ab and
p(±) a , built out of independent zero- and nonzero-modes. The closed-string Poincaré
generators are then [4.10]

J ab = J (+) ab + J (−) ab , pa = p(+) a + p(−) a . (7.1.17a)

Since the operators
∆pa = p(+) a − p(−) a (7.1.17b)

commute with themselves and transform as a vector under the Lorentz algebra, we
can construct a Poincaré algebra from just J ab and ∆pa . This is the Poincaré algebra
whose extension is relevant for string field theory; it closes off shell. (This holds
in either light-cone or covariant quantization.) However, as described above, this

results in an unphysical doubling of zero-modes. This can be fixed by applying the
constraints (see (7.1.10))
∆pa = 0 . (7.1.18)

In the light-cone formalism,
∆p+ = ∆pi = 0 (7.1.19a)

eliminates independent zero-modes, while
1 2
0 = ∆p− = − (M 2(+) − M 2(−) ) = − ∆N (7.1.19b)
p+ p+
is then the usual remaining light-cone constraint equating the numbers of left-handed
and right-handed modes.

The generators of the Lorentz subgroup again take the form (2.3.5), as for the
open string, and the operators appearing in J ab are expressed in terms of the open-
string ones appearing in (7.1.15) as

M ij = M (±) ij ,

M2 = 2 M 2(±) = 4(N − 2) , N= N (±) , ∆N = N (+) − N (−) ,

M im M = 2 (M im M)(±) . (7.1.20)

(Since J ab and pa are expressed as sums, so are M ab and M. This causes objects
quadratic in these operators to be expressed as twice the sums in the presence of the
constraint ∆pa = 0 → p(±) a = 12 pa .)

These Poincaré algebras will be used to derive the OSp(1,1|2) algebras used in
finding gauge-invariant actions in sects. 8.2 and 11.2.

7.2. Spinning

In this section we’ll describe a string model with fermions obtained by introducing
a 2D supersymmetry into the world sheet [7.1,5.1], in analogy to sect. 5.3, and de-
rive the corresponding Poincaré algebra. The description of the superstring obtained
by this method isn’t manifestly spacetime supersymmetric, so we’ll only give a brief
discussion of this formalism before giving the present status of the manifestly super-
symmetric formulation. In both cases, (free) quantum consistency requires D = 10.

For simplicity, we go directly to the (first-)quantized formalism, since as far as
the field theory is concerned the only relevant information from the free theory is how
7.2. Spinning 135

to construct the covariant derivatives from the coordinates, and then the equations of
motion from the covariant derivatives. After quantization, when (in the Schrödinger
picture) the coordinates depend only on σ, there still remains a 1D supersymmetry
in σ-space [7.2]. The covariant derivatives are simply the 1D supersymmetrization of
the P operators of sect. 7.1 (or those of sect. 6.1 at the classical level, using Poisson

{D̂ a (σ 1 , θ1 ), D̂b (σ 2 , θ2 )} = η ab d2 2πδ(σ 2 − σ 1 )δ(θ2 − θ1 ) , (7.2.1a)

D̂(σ, θ) = Ψ̂(σ) + θP (σ)

→ [P , P ] = as bef ore , {Ψ̂a (σ 1 ), Ψ̂b (σ 2 )} = η ab 2πδ(σ 2 − σ 1 ) . (7.2.1b)

The 2D superconformal generators (the supersymmetrization of the generators (6.1.7b)
of 2D general coordinate transformations, or of just the residual conformal transfor-
mations after the lagrange multipliers have been gauged away) are then

2 D̂dD̂ = ( 12 Ψ̂ · P ) + θ( 12 P 2 + 12 iΨ̂ · Ψ̂) . (7.2.2)

There are 2 choices of boundary conditions:

D̂(σ, θ) = ±D̂(σ + 2π, ±θ) → P (σ) = P (σ + 2π) , Ψ̂(σ) = ±Ψ̂(σ + 2π) . (7.2.3)

The + choice gives fermions (the Ramond model), while the − gives bosons (Neveu-

Expanding in modes, we now have, in addition to (7.1.7),

Ψ̂(σ) = γ n e−inσ → {γ a m , γ b n } = η ab δ m+n,0 , γ −n = γ n † , (7.2.4)

where m, n are integral indices for the fermion case and half-(odd)integral for the
bosonic. The assignment of statistics follows from the fact that, while the γ n ’s are

creation operators γ n = dn † for n > 0, they are γ matrices γ 0 = γ/ 2 for n = 0
(as in the particle case, but in relation to the usual γ matrices now have Klein
 or else the
transformation factors for both dn ’s and, in the BRST case, the ghost C,
dn ’s and C are related to the usual by factors of γ 11 ). However, as in the particle
case, a functional analysis shows that this assignment can only be maintained if the
number of anticommuting modes is even; in other words,
d† d
γ 11 (−1) =1 . (7.2.5)

(In terms of the usual γ-matrices, the γ-matrices here also contain the Klein trans-
d† d
formation factor (−1) . However, in practice it’s more convenient to use the
γ-matrices of (7.2.4-5), which anticommute with all fermionic operators, and equate
them directly to the usual matrices after all fermionic oscillators have been elimi-
nated.) As usual, if γ 0 is represented explicitly as matrices, the hermiticity in (7.2.4)
means pseudohermiticity with respect to the time component γ 0 0 (the indefinite met-
ric of the Hilbert space of a Dirac spinor). However, if γ 0 is instead represented as
operators (as, e.g., creation and annihilation operators, as for the usual operator rep-
resentation of SU(N)⊂SO(2N)), no explicit metric is necessary (being automatically
included in the definition of hermitian conjugation for the operators).

The rest is similar to the bosonic formalism, and is straightforward in the 1D
superfield formalism. For example,
dσ dσ 1  2 1 
H= dθ 12 D̂dD̂ = ( P + 2 iΨ̂ Ψ̂) = 12 (p2 + M 2 ) ,
2π 2π 2

M2 = 2 n(an † · an + dn † · dn ) . (7.2.6a)

For the fermionic sector we also have (from (7.2.2))

dσ 1
Ψ̂ · P = √ (/ 
p + M) , 2 = M 2
M . (7.2.6b)
2π 2
(As described above, the d’s anticommute with γ, and thus effectively include an
 is thus analogous to the M
implicit factor of γ 11 . M / of (4.5.12).) The fermionic
ground state is massless (especially due to the above chirality condition), but the
bosonic ground state is a tachyon. (The latter can most easily be seen, as for the
bosonic string, by noting that the first excited level consists of only a massless vector.)
However, consistent quantum interactions require truncation to the spectrum of the
superstring described in the next section. This means, in addition to the chirality
condition (7.2.5) in the fermionic sector, the restriction in the bosonic sector to even
M 2 [7.3]. (Unlike the fermionic sector, odd M 2 is possible because of the half-integral
mode numbers.) As for the bosonic string, besides determining the ground-state
masses Lorentz invariance also fixes the dimension, now D = 10.

In the light-cone formulation of the spinning string we have instead of (7.1.14)

J ab = −iX [a P 0 b] + 12 Ψ̂[a Ψ̂b] ,

X̂ + = p+ σ , Ψ̂+ = 0 ; P 2 + iΨ̂ · Ψ̂ = Ψ̂ · P = 0
7.3. Supersymmetric 137

1  2
→ P − = −
P i + iΨ̂i Ψ̂i , Ψ̂− = − Ψ̂i P i . (7.2.7)
2p+ p+
The resulting component expansion for the Neveu-Schwarz string is similar to the
non-spinning bosonic string, with extra contributions from the new oscillators. For
the case of the Ramond string, comparing to (2.3.5), we find in place of (7.1.15)

M2 = 2 n a† n i ani + d† n i dni ,

M ij = 14 γ [iγ j) + a† n[i anj) + d† n[i dnj) ,

M im M = 1  + 1 γj Σ + Σ j γ +
γ iM a† ni ∆n − ∆† n ani + d† ni Ξn − Ξ† n dni ,
2 2 ji i j

 = Σi
M , Σij = i 2n d† ni anj − dnia† nj ,

1 1 i ) ,
p− = − (pi 2 + M 2 ) , γ− = − (γ pi + M
2p+ p+
1 i
1 i

an− = − p ani − i 12 n i
2 γ dni + ∆n , dn− =− p dni + i n2 γ i ani + Ξn ,
p+ p+
∆n = i √ 2 m(n − m) am an−m,i + (m − i n i
2 )dm dm−n,i
n m=1

† n † i
− m(n + m) a m
an+m,i + (m + 2 )d m dn+m,i ,
 n−1 ∞ √

√ √

† †
Ξn = i i
m am dn−m,i + n+md m
an+m,i − ma m
dn+m,i .
m=1 m=1

This algebra can be applied directly to obtain gauge-invariant actions, as was
described in sect. 4.5.

7.3. Supersymmetric

We now obtain the superstring [7.3] as a combined generalization of the bosonic
string and the superparticle, which was described in sect. 5.4.

Although the superstring can be formulated as a truncation of the spinning
string, a manifestly supersymmetric formulation is expected to have the usual ad-
vantages that superfields have over components in ordinary field theories: simpler
constructions of actions, use of supersymmetric gauges, easier quantum calculations,
no-renormalization theorems which follow directly from analyzing counterterms, etc.
As usual, the free theory can be obtained completely from the covariant derivatives

and equations of motion [7.5]. The covariant derivatives are defined by their affine
Lie, or Kač-Moody (or, as applied to strings, “Kač-Kradle”), algebra of the form
[G i (σ 1 ), G j (σ 2 )} = δ(σ 2 − σ 1 )f ij k G k (σ 1 ) + iδ  (σ 2 − σ 1 )g ij , (7.3.1)

where f are the algebra’s structure constants and g its (not necessarily Cartan) metric
(both constants). The zero-modes of these generators give an ordinary (graded) Lie
algebra with structure constants f . The Jacobi identities are satisfied if and only if

f [ij|l f l|k) m = 0 , (7.3.2a)

f i(j| l g l|k] = 0 , (7.3.2b)

where the first equation is the usual Jacobi identity of a Lie algebra and the second
states the total (graded) antisymmetry of the structure constants with index lowered
by the metric g. In this case, we wish to generalize {dα , dβ } = 2γ a αβ pa for the
superparticle and [P a (σ 1 ), P b (σ 2 )] = 2πiδ  (σ 2 − σ 1 )η ab for the bosonic string. The
simplest generalization consistent with the Jacobi identities is:

{D α (σ 1 ), D β (σ 2 )} = 2πδ(σ 2 − σ 1 )2γ a αβ P a (σ 1 ) ,

[D α (σ 1 ), P a (σ 2 )] = 2πδ(σ 2 − σ 1 )2γ aαβ Ωβ (σ 1 ) ,

{D α (σ 1 ), Ωβ (σ 2 )} = 2πiδ  (σ 2 − σ 1 )δ α β ,

[P a (σ 1 ), P b(σ 2 )] = 2πiδ  (σ 2 − σ 1 )η ab ,

[P, Ω] = {Ω, Ω} = 0 , (7.3.3a)

γ a(αβ γ a γ)δ = 0 . (7.3.3b)

(7.3.2b) requires the introduction of the operator Ω, and (7.3.2a) then implies (7.3.3b).
This supersymmetric set of modes (as P for the bosonic string) describes a complete
open string or half a closed string, so two such sets are needed for the closed super-
string, while the heterotic string needs one of these plus a purely bosonic set.

Note the analogy with the super-Yang-Mills algebra (5.4.8):

(D α , P a , Ωα ) ↔ (∇α , ∇a , W α ) , (7.3.4)

and also that the constraint (7.3.3b) occurs on the γ-matrices, which implies D =
3, 4, 6, or 10 [7.6] when the maximal Lorentz invariance is assumed (i.e., all of SO(D−1,1)
for the D-vector P a ).
7.3. Supersymmetric 139

This algebra can be solved in terms of P a , a spinor coordinate Θα (σ), and its
derivative δ/δΘα :
Dα = α
+ γ a αβ P a Θβ + 12 iγ a αβ γ aγδ Θβ Θγ Θδ ,
Pa = P a + iγ aαβ Θα Θβ ,

Ωα = iΘα . (7.3.5)

These are invariant under supersymmetry generated by
dσ δ
qα = − γ a αβ P a Θβ − 16 iγ a αβ γ aγδ Θβ Θγ Θδ ,
2π δΘα

pa = Pa , (7.3.6)

where {q α , q β } = −2γ a αβ pa .
The smallest (generalized Virasoro) algebra which includes generalizations of the
operators 12 P 2 of the bosonic string and 12 p2 and p/d of the superparticle is generated
A = 12 P 2 + Ωα D α = 12 P 2 + iΘα ,
Bα = γ aαβ P a D β ,

C αβ = 12 D [αD β] ,

D a = iγ a αβ D α D  β . (7.3.7)

Note the similarity of A to (5.4.10), (7.2.6a), and (8.1.10,12). The algebra generated
by these operators is (classically)
[A(1), A(2)] = iδ  (2 − 1)[A(1) + A(2)] ,

[A(1), Bα (2)] = iδ  (2 − 1)[Bα (1) + Bα (2)] ,

[A(1), C αβ (2)] = iδ  (2 − 1)[C αβ (1) + C αβ (2)] ,

[A(1), D a (2)] = iδ  (2 − 1)[D a (1) + 2D a (2)] ,

{Bα (1), Bβ (2)} = iδ  (2 − 1) 12 γ aαγ γ a βδ [C γδ (1) + C γδ (2)] + 4δ(2 − 1)·

· [γ aαβ (P a A + 18 D a ) + (δ γ (α δ δ β) − 12 γ a αβ γ a γδ )Ωγ Bδ ] ,
1 α
[B (1), C βγ (2)] = 4δ(2 − 1)[δ [β α D γ] A + (δ δ  δ [β α − 12 γ a α γ a δ[β )Ωδ C γ] ] ,


1 α
[B (1), D a (2)] = − 2iδ  (2 − 1)[2γ a αβ D β A + (3δ γ  γ a αδ − γ abγ  γ bαδ )·

· Ωγ C δ ](1) + 2iδ(2 − 1)[4γ a αβ D  β A

+ (3δ γ  γ a αδ − γ abγ  γ bαδ )Ωγ C  δ − iγ aβγ γ bαγ Ωβ D b ] ,
[C αβ (1), C γδ (2)] = − 2δ(2 − 1)P a γ a [γ[α C β]δ] ,

[C αβ (1), D a (2)] = 2iδ  (2 − 1)γ a γδ γ bγ[α P b C β]δ (1)

− 4iδ(2 − 1)(γ aγ[α D  β] Bγ + P b γ ba[α δ C  β]δ ) ,
[D a (1), D b (2)] = − δ  (2 − 1)γ abα β D β Bα (1) + (2)

+ δ  (2 − 1) −4iP (a D b) + η ab (3D  α Bα − D α Bα ) (1) − (2)

+ δ(2 − 1) −2i(3P  [a D b] − P [a D  b] )

+2γ abα β (3D  β Bα − D  β Bα ) + γ abc αβ (3P c C  αβ − P c C αβ ) .


(Due to identities like P/ DD ∼ BD ∼ P/ C, there are other forms of some of these

BRST quantization can again be performed, and there are an infinite number of
ghosts, as in the particle case. However, a remaining problem is to find the appropri-
ate ground state (and corresponding string field). Considering the results of sect. 5.4,
this may require modification of the generators (7.3.7) and BRST operator, perhaps
to include Lorentz generators (acting on the ends of the string?) or separate contri-
butions from the BRST transformations of Yang-Mills field theory (the ground state
of the open superstring, or of a set of modes of one handedness of the corresponding
closed strings). On the other hand, the ground state, rather than being Yang-Mills,
might be purely gauge degrees of freedom, with Yang-Mills appearing at some excited
level, so modification would be unnecessary. The condition Q2 = 0 should reproduce
the conditions D = 10, α0 = 1.

The covariant derivatives and constraints can also be derived from a 2D lagrangian
of the general form (3.1.10), as for the superparticle [7.5]. This classical mechanics
Lagrangian imposes weaker constraints than the Green-Schwarz one [7.6] (which sets
D α = 0 via Gupta-Bleuler), and thus should not impose stronger conditions.

On the other hand, quantization in the light-cone formalism is understood. Spinors
are separated into halves, with the corresponding separation of the γ matrices giving
7.3. Supersymmetric 141

the splitting of vectors into transverse and longitudinal parts, as in (5.4.27). The
light-cone gauge is then chosen as

P + (σ) = p+ , Θ− = 0 . (7.3.9)

Other operators are then eliminated by auxiliary field equations:
1 1  2 +

A =0 → P− = − 2 P i + iΘ D + ,
1  †
B =0 → D− = √ P/ T D + . (7.3.10)

The remaining coordinates are x± , X i (σ), and Θ+ (σ), and the remaining operators
D+ = +
+ p+ ΣΘ+ , P i = P i , Ω+ = iΘ+ . (7.3.11)
However, instead of imposing C = D = 0 quantum mechanically, we can solve them
classically, in analogy to the particle case. The C +ij (σ) are now local (in σ) SO(8)
generators, and can be used to gauge away all but 1 Lorentz component of D + , by
the same method as (5.4.34ab) [5.30,29]. After this, D + is just the product of this
one component times its σ-derivative. Furthermore, D + is a Virasoro algebra for D + ,
and can thus be used to gauge away all but the zero-mode [5.30,29] (as the usual one
A did for P + in (7.3.9)), after this constraint implies D + factors in a way analogous
to (5.4.34a):
D+ = 0 → D + = cξ(σ) . (7.3.12)

(The proof is identical, since D + = 0 is equivalent to D + (σ 1 )D + (σ 2 ) = 0.) We
are thus back to the particle case for D, with a single mode remaining, satisfying

the commutation relation c2 = p+ → c = ± p+ , so D is completely determined.
Alternatively, as for the particle, we could consider D = 0 as a second-class constraint
[7.6], or impose the condition D + = 0 (which eliminates all auxiliary string fields),
as a Gupta-Bleuler constraint. This requires a further splitting of the spinors, as in
(5.4.29), and the Gupta-Bleuler constraint is again a chirality condition, as in (5.4.33).
C = D = 0 are then also satisfied ala Gupta-Bleuler (with appropriate “normal
ordering”). Thus, in a “chiral” representation (as in ordinary supersymmetry) we have
a chiral, “on-shell” string superfield, or wave function, Φ[x± , X i(σ), Θa (σ)], which
satisfies a light-cone field equation

i +H Φ=0 ,

1 dσ  dσ 1  2 δ
H = −p− = −  P− = 2P i + iΘ a
. (7.3.13)
α 2π 2π δΘa
The dimension of spacetime D = 10 and the constant in H (zero) are determined by
considerations similar to those of the bosonic case (Lorentz invariance in the light-cone
formalism, or BRST invariance in the yet-to-be-constructed covariant formalism).

Similarly, light-cone expressions for q α can be obtained from the covariant ones:
dσ δ + dσ
q+ = +
− p + ΣΘ ≡ Q+ ,
2π δΘ 2π

dσ 1  †
q− = √ P/ T Q+ . (7.3.14)
2π 2p+

If the superstring is formulated directly in the light cone, (7.3.14) can be used as the
starting point. Q+ and D + can be considered as independent variables (instead of
Θ+ and δ/δΘ+ ), defined by their self-conjugate commutation relations (analogous to
those of P ):
{Q+ (σ 1 ), Q+ (σ 2 )} = −δ(σ 2 − σ 1 )2p+ Σ ,

{D + (σ 1 ), D + (σ 2 )} = δ(σ 2 − σ 1 )2p+ Σ . (7.3.15)

However, as described above and for the particle, D + is unnecessary for describing
physical polarizations, so we need not introduce it. In order to more closely study
the closure of the algebra (7.3.14), we introduce more light-cone spinor notation (see
sect. 5.3): Working in the Majorana representation Σ = I, we introduce (D − 2)-
dimensional Euclidean γ-matrices as

p/T → γ i µν  pi ,

γ (i µµ γ j) νµ = 2δ ij δ µν , γ (i µµ γ j) µν  = 2δ ij δ µ ν  , (7.3.16)

where not only vector indices i but also spinor indices µ and µ can be raised and
lowered by Kronecker δ’s, and primed and unprimed spinor indices are not necessarily
related. (However, as for the covariant indices, there may be additional relations
satisfied by the spinors, irrelevant for the present considerations, that differ in different
dimensions.) Closure of the supersymmetry algebra (on the momentum in the usual
way (5.4.4), but in light-cone notation, and with the light-cone expressions for pi , p+ ,
and p− ) then requires the identity (related to (7.3.16) by “triality”)

γ i(µ|µ γ i |ν)ν  = 2δ µν δ µ ν  . (7.3.17)
7.3. Supersymmetric 143

This identity is actually (7.3.3b) in light-cone notation, and the equality of the di-
mensions of the spinor and vector can be derived by tracing (7.3.17) with δ µν .

Returning to deriving the light-cone formalism from the covariant one, we can
also obtain the light-cone expressions for the Poincaré generators, which should prove
important for covariant quantization, via the OSp(1,1|2) method. As in general, they
are completely specified by M ij , M im M, and M 2 :
M ij = −iX̂ i P j + Mij ,
M im M = iX̂ i p+ P − + Mi j P j ,
M2 = −2p+ P − , (7.3.18)

1 1
−p+ P − = 12 P i 2 + i
Qγ + Q − i D+ ,
8p+ 8p+
1 1
Mij = − Qγ + γ ij Q + C +ij , (7.3.19)
16p+ 16p+
contain all D dependence (as opposed to X and Q dependence) only in the form of
C and D, which can therefore be dropped. (Cf. (5.4.22). γ + picks out Q+ from Q, as
in (5.4.27).)

We can now consider deriving the BRST algebra by the method of adding 4+4
dimensions to the light-cone (sects. 3.6, 5.5). Unfortunately, adding 4+4 dimensions
doesn’t preserve (7.3.17). In fact, from the analysis of sect. 5.3, we see that to pre-
serve the symmetries of the σ-matrices requires increasing the number of commuting
dimensions by a multiple of 8, and the number of time dimensions by a multiple of 4.
This suggests that this formalism may need to be generalized to adding 8+8 dimen-
sions to the light-cone (4 space, 4 time, 8 fermionic). Coincidentally, the light-cone
superstring has 8+8 physical (σ-dependent) coordinates, so this would just double
the number of oscillators. Performing the reduction from OSp(4,4|8) to OSp(2,2|4)
to OSp(1,1|2), if one step is chosen to be U(1)-type and the other GL(1)-type, it may
be possible to obtain an algebra which has the benefits of both formalisms.

As for the bosonic string, the closed superstring is constructed as the direct
product of 2 open strings (1 for the clockwise modes and 1 for the counterclockwise):
The hamiltonian is the sum of 2 open-string ones, and the closed-string ground state
is the product of 2 open-string ones. In the case of type I or IIB closed strings, the
2 sets of modes are the same kind, and the former (the bound-state of type I open

strings) is nonoriented (in order to be consistent with the N = 1 supersymmetry of
the open string, rather than the N = 2 supersymmetry generated by the 2 sets of
modes of oriented, type II strings). Type IIA closed strings have Θ’s with the opposite
chirality between the two sets of modes (i.e., one set has a Θα while the other has a
Θα ). The ground states of these closed strings are supergravity (N = 1 supergravity
for type I and N = 2 for type II). The heterotic string is a closed string for which
one set of modes is bosonic (with the usual tachyonic scalar ground state) while the
other is supersymmetric (with the usual supersymmetric Yang-Mills ground state).
The lowest-mass physical states, due to the ∆N = 0 restriction, are the product
of the massless sector of each set (since now ∆N = H (+) − H (−) = (N (+) − 1) −
N (−) ). The dimension of spacetime for the 2 sets of modes is made consistent by
compactification of some of the 26 dimensions of one and some (or none) of the 10 of
the other onto a torus, leaving the same number of noncompactified dimensions (at
least the physical 4) for both sets of modes. These compactified bosonic modes can
also be fermionized (see the next section), giving an equivalent formulation in which
the extra dimensions don’t explicitly appear: For example, fermionization of 16 of
the dimensions produces 32 (real) fermionic coordinates, giving an SO(32) internal
symmetry (when the fermions are given the same boundary conditions, all periodic or
all antiperiodic). The resulting spectrum for the massless sector of heterotic strings
consists of supergravity coupled to supersymmetric Yang-Mills with N = 1 (in 10D
counting) supersymmetry. The vectors gauging the Cartan subalgebra of the full
Yang-Mills group are the obvious ones coming from the toroidal compactification
(i.e., those that would be obtained from the noncompactified theory by just dropping
dependence on the compactified coordinates), while the rest correspond to “soliton”
modes of the compactified coordinates for which the string winds around the torus. As
for the dimension and ground-state mass, quantum consistency restricts the allowed
compactifications, and in particular the toroidal compactifications are restricted to
those which, in the case of compactification to D = 10, give Yang-Mills group SO(32)
or E8 ⊗E8 . (These groups give anomaly-free 10D theories in their massless sectors.
There is also an SO(16)×SO(16) 10D-compactification which can be considered to
have broken N=1 supersymmetry. There are other 10D-compactifications which have

Some aspects of the interacting theory will be described in chapts. 9 and 10.
Exercises 145


(1) Use (7.1.7) as the classical solution for X̂ i , and set ani = 0 for n = 1 and i = 1.
Find X a (σ, τ ). In the center-of-mass frame, find the energy and spin, and relate

(2) Do (1) in the conformal gauge by using (7.1.7) for X̂ a for a = 0, 1 (same n), and
applying the constraint (6.2.5). Compare results.

(3) Prove the light-cone Poincaré algebra closes only for D = 26, and determines the
constant in (7.1.8a).

(4) Find explicit expressions for all the states at the 4 lowest mass levels of the open
bosonic string. For the massive levels, combine SO(D−2) representations into
SO(D−1) ones. Do the same for the 4 lowest (nontrivial) mass levels of the
closed string.

(5) Derive (7.1.15), including the expressions in terms of σ-integrals. What happens
to the part of this integral symmetric in ij for M ij ?

(6) Derive (7.2.8).

(7) Derive (7.3.17), both from (7.3.3b) and closure of (7.3.14). Show that it implies
D − 2 = 1, 2, 4, 8.

(8) Show that (7.3.5) satisfies (7.3.3). Show that (7.3.6) gives a supersymmetry
algebra, and that the operators of (7.3.5) are invariant. Check (7.3.8) till you

(9) Find all the states in the spinning string at the tachyonic, massless, and first
massive levels. Show that, using the truncation of sect. 7.2, there are equal
numbers of bosons and fermions at each level. Construct the same states using
the X and Q oscillators of sect. 7.3.


8.1. IGL(1)

We first describe the form of the BRST algebra obtained by first-quantization
of the bosonic string by the method of sect. 3.2, using the constraints found in the
conformal (temporal) gauge in sect. 6.2.

The residual gauge invariance in the covariant gauge is conformal transformations
(modified by the constraint that they preserve the position of the boundaries). After
quantization in the Schrödinger picture (where the coordinates have no τ dependence),
the Virasoro operators [8.1]

1 2
G(σ) = −i 2 P (σ) −1 , (8.1.1)

with P as in (7.1.3) but for all Lorentz components, generate only these transfor-
mations (instead of the complete set of 2D general coordinate transformations they
generated when left as arbitrary off-shell functions of σ and τ in the classical mechan-
ics). Using the hamiltonian form of BRST quantization, we first find the classical
commutation relations (Poisson brackets, neglecting the normal-ordering constant in

[G(σ 1 ), G(σ 2 )] = 2πδ  (σ 2 − σ 1 )[G(σ 1 ) + G(σ 2 )]

= 2π [δ(σ 2 − σ 1 )G  (σ 2 ) + 2δ  (σ 2 − σ 1 )G(σ 2 )] , (8.1.2a)

or in mode form

iG(σ) = Ln e−inσ →

[Lm , Ln ] = (n − m)Lm+n . (8.1.2b)

These commutation relations, rewritten as
dσ 1 dσ 2 dσ
λ1 (σ 1 )G(σ 1 ), λ2 (σ 2 )G(σ 2 ) = λ[2 (σ)λ1]  (σ)G(σ) , (8.1.3a)
2π 2π 2π
8.1. IGL(1) 147

correspond to the 1D general coordinate transformations (in “zeroth-quantized” no-
∂ ∂ ∂
λ2 (σ) , λ1 (σ) = λ[2 λ1]  , (8.1.3b)
∂σ ∂σ ∂σ
giving the structure constants

f (σ 1 , σ 2 ; σ 3 ) = 2πδ  (σ 2 − σ 1 ) [δ(σ 1 − σ 3 ) + δ(σ 2 − σ 3 )] , (8.1.4a)


λ1 λ2 f ij j
↔ λ1 λ2 .
i k
More generally, we have operators whose commutation relations

[G(σ 1 ), O(σ 2 )] = 2π [δ(σ 2 − σ 1 )O (σ 2 ) + wδ (σ 2 − σ 1 )O(σ 2 )] (8.1.5a)

represent the transformation properties of a 1D tensor of (covariant) rank w, or a
scalar density of weight w:
λG, O = λO + wλO . (8.1.5b)

Equivalently, in terms of 2D conformal transformations, it has scale weight w. (Re-
member that conformal transformations in D = 2 are equivalent to 1D general co-
ordinate transformations on σ ± : See (6.2.7).) In particular, we see from (8.1.2a)
that G itself is a 2nd-rank-covariant (as opposed to contravariant) tensor: It is the
energy-momentum tensor of the mechanics action. (It was derived by varying that
action with respect to the metric.) The finite form of these transformations follows
from exponentiating the Lie algebra represented in (8.1.3): (8.1.5) can then also be
rewritten as the usual coordinate transformations
O (σ  ) = O(σ) , (8.1.6a)

where the primes here stand for the transformed quantities (not σ-derivatives) or as

(dσ  )w O (σ  ) = (dσ)w O(σ) , (8.1.6b)

indicating their tensor structure. In particular, a covariant vector (w = 1) can be
integrated to give an invariant. P is such a vector (and the momentum p the corre-
sponding conformal invariant), which is why G has twice its weight (by (8.1.1)).

Before performing the BRST quantization of this algebra, we relate it to the
light-cone quantization of the previous chapter. The constraints (8.1.1) can be solved

in a Gupta-Bleuler fashion in light-cone notation. The difference between that and
actual light-cone quantization is that in the light-cone quantization P − is totally
eliminated at the classical level, whereas in the light-cone notation for the covariant-
gauge quantization the constraint is used to determine the dependence on the ±
oscillators in terms of the transverse oscillators. One way to do this would be to start
with a state constructed from just transverse oscillators (as in light-cone quantization)
and add in terms involving longitudinal oscillators until the constraints are satisfied
(or actually half of them, ala Gupta-Bleuler). A simpler way is to start at the classical
level in an arbitrary conformal gauge with transverse oscillators, and then conformally
transform them to the light-cone gauge to see what a transverse oscillator (in the
physical sense, not the light-cone-index sense) looks like. We thus wish to consider

dσ  P i  (σ  ) = dσ P i (σ) ,
σ = X̂ + (σ) = σ + oscillator-terms . (8.1.7)
(Without loss of generality, we can work at x+ = 0. Equivalently, we can explicitly
subtract x+ from X̂ + everywhere the latter appears in this derivation.) If we consider
the same transformation on P + (using ∂ X̂/∂σ ≡ P ), we find P +  = p+ , the light-cone
gauge. (8.1.7) can be rewritten as
P  (σ 1 ) = dσ 2 δ σ 1 − X̂ + (σ 2 ) P (σ 2 ) . (8.1.8)
(8.1.7) follows upon replacing σ 1 with σ 1  and integrating out the δ-function (with
the Jacobian giving the conformal weight factor). A more convenient form for quan-
tization comes from the mode expansion: Multiplying by einσ1 and integrating,

 dσ inX̂ + (σ)/p+ 
αn = e P (σ) . (8.1.9)

These (“DDF”) operators [8.2] (with normal ordering, as usual, upon quantization,
and with transverse Lorentz index i, and n > 0) can be used to create all physi-
cal states. Due to their definition in terms of a conformal transformation from an
arbitrary conformal gauge to a completely fixed (light-cone) gauge, they are auto-
matically conformally invariant: i.e., they commute with G. (This can be verified to
remain true after quantization.) Consequently, states constructed from them satisfy
the Gupta-Bleuler constraints, since the conformal generators push past these opera-
tors to hit the vacuum. Thus, these operators allow the construction of the physical
Hilbert space within the formalism of covariant-gauge quantization, and allow a direct
comparison with light-cone quantization.
8.1. IGL(1) 149

On the other hand, for most purposes it is more convenient to solve the constraints
as covariantly as possible (which is we why we are working with covariant-gauge
quantization in the first place). The next step is the IGL(1) algebra [3.4]
dσ  1 2   δ dσ 
Q= C −i 2 P + C  + i ∝ 0 ≡ −i CA ,
2π δC 2π

dσ  δ
J3 = C . (8.1.10)
2π δ C
Expanding in the ghost zero-mode

c= C (8.1.11)

we also find (see (3.4.3b))
p +M =2 2
A , +
M = −i C C  . (8.1.12)

∝ 0 (the intercept of the leading Regge trajectory) is a constant introduced, as in
the light-cone formalism, because of implicit normal ordering. The only ambiguous
constant in J 3 is an overall one, which we choose to absorb into the zero-mode term so
that it appears as c∂/∂c, so that physical fields have vanishing ghost number. (This
also makes J 3† = 1 − J 3 .) In analogy to the particle, C is a momentum (defined to
be periodic on σ ∈ [−π, π]), as follows from consideration of τ reversal in the classical
mechanics action, but here τ reversal is accompanied by σ reversal in order to avoid
switching + and − modes. (In the classical action the ghost is odd under such a
transformation, since it carries a 2D vector index, as does the gauge parameter, while
the antighost is even, carrying 2 indices, as does the gauge-fixing function gmn .) C can
also be separated into odd and even parts, which is useful when similarly separating
P as in (7.1.3a):

C = C + C , C(−σ) = C(σ) , 
= −C(σ) . (8.1.13)

We now pay attention to the quantum effects. Rather than examining the BRST
algebra, we look at the IGL(1)-invariant Virasoro operators (from (3.2.13))
δ δ
Ĝ = G + C  
+ C  + i(∝ 0 − 1) . (8.1.14)
δC δC
(The ∝ 0 − 1 just replaces the 1 in (8.1.1) with ∝ 0 .) Corresponding to (8.1.2b),
we now have the exact quantum mechanical commutation relations (after normal
D − 26 3
[L̂m , L̂n ] = (n − m)L̂m+n + (m − m) + 2(∝ 0 − 1)m δ m,−n . (8.1.15)

The terms linear in m in the anomalous terms (those not appearing in the classical
result (8.1.2)) are trivial, and can be arbitrarily modified by adding a constant to L̂0 .
That the remaining term is ∼ m3 follows from (1D) dimensional analysis: [Ĝ, Ĝ] ∼
δ  Ĝ + δ  , since the first term implies Ĝ ∼ 1/σ 2 dimensionally, so only δ  ∼ 1/σ 4
can be used. The values of the coefficients in these terms can also be determined
by evaluating just the vacuum matrix elements 0| [L̂−n , L̂n ] |0
for n = 1, 2. Further
examining these terms, we see that the ghost contributions are necessary to cancel
those from the physical coordinates (which have coefficient D), and do so only for
D = 26. The remaining anomaly cancels for ∝ 0 = 1. Under the same conditions one
can show that Q2 = 0. Thus, in the covariant formalism, where Lorentz covariance is
manifest and not unitarity (the opposite of the light-cone formalism), Q2 = 0 is the
analog of the light cone’s [J −i , J −j ] = 0 (and the calculation is almost identical, so the
reader will have little trouble modifying his previous calculation). C and δ/δ C can be
expanded in zero-modes and creation and annihilation operators, as P ((7.1.7a)), but
the creation operators in C are canonically conjugate to the annihilation operators in
 and vice versa:
δ/δ C,

C = c + √ (cn † e−inσ + cn einσ ) ,
1 n
∞ √

δ ∂
= + n(−ic̃n † e−inσ + ic̃n einσ ) ;
δ C ∂c 1

{cm , c̃n † } = iδ mn , {c̃m , cn † } = −iδ mn . (8.1.16)
(Since the IGL(1) formalism is directly related to the OSp(1,1|2), as in sect. 4.2, we
have normalized the oscillators in a way that will make the Sp(2) symmetry manifest
in the next section.) The physical states are obtained by hitting |0
with a† ’s but also
requiring Q |ψ
= 0; states |ψ
= Q |χ
are null states (pure gauge). The condition
of being annihilated by Q is equivalent to being annihilated by Ln for n ≤ 0 (i.e.,
the “nonpositive energy” part of G(σ), which is now normal ordered and includes the
− ∝ 0 term of (8.1.10,14)), which is just the constraint in Gupta-Bleuler quantization.
L0 is simply the Lorentz-covariantization of H of (7.1.8) (i.e., all transverse indices
replaced with Lorentz indices).
An interesting fact about the Virasoro algebra (8.1.15) (and its generalizations,
see below) is that, after an appropriate shift in L0 (namely, the choice of ∝ 0 = 1 in
this case), the anomaly does not appear in the Sp(2) (=SL(2)=SU(1,1)=SO(2,1)=
projective group) subalgebra given by n = 0, ±1 [8.3], independent of the represen-
tation (in this case, D). Furthermore, unlike the whole Virasoro algebra (even when
8.1. IGL(1) 151

the anomaly cancels), we can define a state which is left invariant by this Sp(2).
Expanding (8.1.14) in modes (as in (7.1.7a, 8.1.2b)), the only term in L̂1 containing
no annihilation operators is ∼ p · a1 † , so we choose p = 0. Then L̂0 = 0 requires the
state be on-shell, which means it’s in the usual massless sector (Yang-Mills). Further
examination then shows that this state is uniquely determined to be the state corre-
sponding to a constant (p = 0) Yang-Mills ghost field C. It can also be shown that
this is the only gauge-invariant, BRST-invariant state (i.e., in the “cohomology” of
Q) of that ghost-number J 3 [8.4]. Since it has the same ghost number as the gauge
parameter Λ (see (4.2.1)), this means that it can be identified as the only gauge in-
variance of the theory which has no inhomogeneous term: Any gauge parameter of
the form Λ = Q
not only leaves the free action invariant, but also the interacting one,
since upon gauge fixing it’s a gauge invariance of the ghosts (which means the ghosts
themselves require ghosts), which must be maintained at the interacting level for
consistent quantization. However, any parameter satisfying QΛ = 0 won’t contribute
to the free gauge transformation of the physical fields, but may contribute at the
interacting level. In fact, gauge transformations in the cohomology of Q are just the
global invariances of the theory, or at least those which preserve the second-quantized
vacuum about which the decomposition into free and interacting has been defined.
Since the BRST transformation δΦ = QΦ is just the gauge transformation with the
gauge parameter replaced by the ghost, this transformation parameter appears in the
field in the same position as would the corresponding ghost. For the bosonic string,
the only massless physical field is Yang-Mills, and thus the only global invariance is
the usual global nonabelian symmetry. Thus, the state invariant under this Sp(2) di-
rectly corresponds to the global invariance of the string theory, and to its ghost. This
Sp(2) symmetry can be maintained at the interacting level in tree graph calculations
(see sect. 9.2), especially for vertices, basically due to the fact that tree graphs have
the same global topology as free strings. In such calculations it’s therefore somewhat
more convenient to expand states about this Sp(2)-invariant “vacuum” instead of the
usual one. (We now refer to the first-quantized vacuum with respect to which free
fields are defined. It’s redefinition is unrelated to the usual vacuum redefinitions of
field theory, which are inhomogeneous in the fields.) This effectively switches the role
of the corresponding pair of ghost oscillators (just the n = 1 mode) between creation
and annihilation operators.

The closed string [4.5] is quantized similarly, but with 2 sets of modes (±; except

that there are still just one x and p), and we can separate

C ± (σ) = (C ± C)(±σ)

corresponding to (7.1.6).

Since A commutes with both M 3 and M + , it is Sp(2)-invariant. Thus, the mod-
ified Virasoro operators Ľn it gives (in analogy to (8.1.2b), or, more specifically, the
nonnegative-energy ones), and in particular their fermionic parts, can be used to
generate (BRST) Sp(2)-invariant states, with the exception of the zeroth and first
fermionic Virasoro operators (the projective subgroup), which vanish on the vacuum.
We will now show that these operators, together with the bosonic oscillators, are
sufficient to generate all such states, i.e., the complete set of physical fields [4.1]. (By
physical fields we mean all fields appearing in the gauge-invariant action, including
Stueckelberg fields and unphysical Lorentz components.) This is seen by bosonizing
the two fermionic coordinates into a single additional bosonic coordinate, whose con-
tribution to the Virasoro operators includes a term linear in the new oscillators, but
lacking the first mode. This corresponds to the fact that M + contains a term linear
in the annihilation operator of the first mode. Thus, the Virasoro operators generate
excitations in all but the first mode of the new coordinate, and the condition M + = 0
kills only excitations in the first mode. J 3 is just the zero mode of the new coordinate,
so its vanishing (which then implies T = 0) completes the derivation.

The bosonization is essentially the same as the standard procedure [8.5], except
for differences due to the indefinite metric of the Hilbert space of the ghosts. The
fermionic coordinates can be expressed in terms of a bosonic coordinate χ̂ (analogous
to X̂) as
C = eχ̂ 
= e−χ̂ ,
, (8.1.18)
with our usual implicit normal ordering (with both terms q̂ + p̂σ of the zero mode
appearing in the same exponential factor). Note the hermiticity of these fermionic
coordinates, due to the lack of i’s in the exponents. (For physical bosons and fermions,
we would use ψ̂ = eiϕ̂ , ψ̂ † = e−iϕ̂ , with ψ̂ canonically conjugate to ψ̂ † .) χ̂ has the
mode expansion

χ̂ = (q̂ + p̂σ) + √ (ân einσ + ↠n e−inσ ) ;
n=1 n

[p̂, q̂] = −i , [âm , ↠n ] = −δ mn . (8.1.19)
8.1. IGL(1) 153

By comparison with (7.1.7), we see that this coordinate has a timelike metric (i.e.,
it’s a ghost). Using

: eaχ̂(σ) : : ebχ̂(σ ) : = : eaχ̂(σ)+bχ̂(σ ) : 2i sin σ 2−σ +
, (8.1.20)

we can verify the fermionic anticommutation relations, as well as

3 1 + dσ 2χ̂
J = ip̂ + 2 , M = e ; (8.1.21a)

A= 1
2 P 2 − χ̂2 − χ̂ − 9
4 . (8.1.21b)

Since J 3 is quantized in integral values, χ is defined to exist on a circle of imaginary
length with anticyclic boundary conditions. (The imaginary eigenvalues of this hermi-
tian operator are due to the indefiniteness induced by the ghosts into the Hilbert-space
metric.) Conversely, choosing such values for ip̂ makes C periodic in σ. The SU(2)
which follows from J 3 and M + is not the usual one constructed in bosonization [8.6]
because of the extra factors and inverses of ∂/∂σ involved (see the next section).

Since we project onto p̂ = i 12 when acting on Φ, we find for the parts of M + and
A linear in χ oscillators when acting on Φ

M + = e2q̂ 2â1 + · · · , Ľn = 1
2 n(n − 1)↠n + · · · . (8.1.22)

This shows how the constraint T = 0 essentially just eliminates the zeroth and first
oscillators of χ.

We have seen some examples above of Virasoro operators defined as expressions
quadratic in functions of σ (and their functional derivatives). More generally, we
can consider a bosonic (periodic) function fˆ(σ) with arbitrary weight w. In order to
obtain the transformation law (8.1.5), we must have
δ δ
G(σ) = fˆ − w fˆ , (8.1.23a)
δfˆ δ fˆ
up to an overall normal-ordering constant (which we drop). By manipulations like
those above, we find
[Lm , Ln ] = (n − m)Lm+n + (w − 12 )2 − 1
12 m3 − 16 m δ m,−n . (8.1.23b)

Since fˆ and δ/δ fˆ (or fˆ and −δ/δ fˆ ) have the same commutation relations as two P ’s,
but with off-diagonal metric η ab = ( 01 10 ), for w = 0 (or w = 1) the algebra (8.1.23)
must give just twice the contribution to the anomaly as a single P . This agrees exactly

ˆ the anomalous terms in (8.1.23) have the
with (8.1.15) (the D term). For fermionic f,
opposite overall sign. In that case, fˆ and δ/δ fˆ have the anticommutation relations
of 2 physical fermions (see sect. 7.2), again with the off-diagonal metric, and w = 2
gives the Virasoro operators for 2 physical fermions (i.e., as in the above bosonic
case, G can be rewritten as the sum of 2 independent G’s). The anomaly for a single
physical fermion in thus given by half of that in (8.1.23b), with opposite sign. Another
interesting case is w = −1 (or 2), which, for fermions, gives the ghost contribution
of (8.1.15) (the non-D terms, for ∝ 0 = 0; comparing (8.1.14) with (8.1.23a), we see
C has w = −1 and thus δ/δ C has w = 2). Thus, (8.1.23) is sufficient to give all the
Virasoro algebras which are homogeneous of second order in 1D functions. By the
method of bosonization (8.1.18), the fermionic case of (8.1.23a) can be rewritten as

1 2
G=i 2P + ( 12 − w)P  + 1
8 , (8.1.23c)

where fˆ = exp χ̂ in terms of a timelike coordinate χ (P = χ̂ ). For w = 12 , this
gives an independent demonstration that 2 physical fermions give the same anomaly
as 1 physical boson (modulo the normal-ordering constant), since they are physically
equivalent (up to the boundary conditions on the zero-modes). (There are also factors
of i that need to be inserted in various places to distinguish physical bosons and
fermions from ghost ones, but these don’t affect the value of the anomaly.)

As before, these Virasoro operators correspond to 2D energy-momentum tensors
obtained by varying an action with respect to the 2D metric. Using the vielbein
formalism of sect. 4.1, we first note that the Lorentz group has only one generator,
which acts very simply on the light-cone components of a covariant tensor:

M ab =
ab M (
+− = 1) → [M, ψ (s) ] = sψ (s) , (8.1.24)

where for tensors s is the number of “+” indices minus “−” indices. However, since
the 2D Lorentz group is abelian, this generalizes to arbitrary “spin,” half-integral as
well as irrational. The covariant derivative can then be written as

∇a = ea + ω a M ,

ω a = 12
cb ω abc = 12
cb cbca = −
ab e∂ m e−1 ebm . (8.1.25)

We also have the only nonvanishing component of the curvature Rabcd given by

e−1 R =
mn ∂ n ω m = ∂ m eam ∂ n (e−1 ea n ) . (8.1.26)
8.1. IGL(1) 155

The covariant action corresponding to (8.1.23a) is

S∼ d2 x e−1 (ψ (+) 1−w ∇− ψ (+) w + ψ (−) 1−w ∇+ ψ (−) w ) ,

[M, ψ (±) w ] = ±wψ (±) w , (8.1.27)

where ψ (±) w corresponds to fˆ and ψ (±) 1−w to δ/δ f.
ˆ For open-string boundary con-
ditions, ψ (±) w are combined to form fˆ (as, for w = 1, P ± combined to form P in
(7.1.3a)); for closed strings, the 2 functions can be used independently (as the usual
(±) modes for closed strings). We thus see that the spin is related to the weight (at
least for these free, classical fields) as s = ±w for ψ (±) w . The action corresponding
to (8.1.23c) (neglecting the normal-ordering constant) is [8.7]

S∼ d2 x e−1 1
2χ 2 χ + (w − 12 )Rχ . (8.1.28a)

(We have dropped some surface terms, as in (4.1.36).) The fact that (8.1.28a) repre-
sents a particle with spin can be seen in (at least) 2 ways. One way is to perform a
duality transformation [8.8]: (8.1.28a) can be written in first-order form as

S∼ d2 x e−1 1
2 (F a )
+ F a ∇a χ + (w − 12 )Rχ . (8.1.28b)

(Note that ∇χ is the field strength for χ under the global invariance χ → χ + constant.
In that respect, the last term in (8.1.28b) is like a “Chern-Simons” term, since it can
only be written as the product of 1 field with 1 field strength, in terms of the fields
ω a and χ and their field strengths R and ∇a χ.) Eliminating F by its equation of
motion gives back (8.1.28a), while eliminating χ gives

S∼ d2 x e−1 12 (Ga )2 ,

Ga = −
ab ∇b φ , [M, φ] = w − 2 . (8.1.28c)

(Actually, since (8.1.28a) and (8.1.28c) are equivalent on shell, we could equally well
have started with (8.1.28c) and avoided this discussion of duality transformations.
However, (8.1.28a) is a little more conventional-looking, and the one that more com-
monly appears in the literature.) The unusual Lorentz transformation law of φ follows
from the fact that it’s the logarithm of a tensor:

[M, eφ ] = (w − 12 )eφ , [M, e−φ ] = ( 12 − w)e−φ . (8.1.29)

This is analogous to (8.1.18), but the weights there are increased by 2 by quantum
effects. (More examples of this effect will be discussed in sect. 9.1.)

Another way to see that χ has an effective Lorentz weight w is to look at the
relationship between Lorentz weights and weights under 2D general coordinate trans-
formations (or 1D, or 2D conformal transformations), as in (8.1.5). This follows
from the fact that conformally invariant theories, when coupled to gravity, become
locally scale invariant theories (even without introducing the scale compensator φ
of (4.1.34)). (Conversely, conformal transformations can be defined as the subgroup
of general coordinate + local scale transformations which leaves the vacuum invari-
ant.) This means that we can gauge-transform e to 1, or, equivalently, redefine the
nongravitational fields to cancel all dependence on e. Then ea m appears only in the
unit-determinant combination ea m = e−1/2 ea m . The weights w then appear in the
scale transformation which leaves (8.1.27) invariant:

ψ w  = ewζ ψ w . (8.1.30)

(This has the same form as a local Lorentz transformation, but with different relative
signs for the fields in the (±) terms of (8.1.27). This is related to the fact that, upon
applying the equations of motion, and in the conformal gauge, the 2 sets of fields de-
pend respectively on τ ∓σ, and therefore have independent conformal transformations
on these ± coordinates, except as related by boundary conditions for the open string.)
Choosing eζ = e−1/2 then replaces the ψ’s with fields which are scale-invariant, but
transform under general coordinate transformations as densities of weight w (i.e., as
tensors times e−w/2 ). In the conformal gauge, these densities satisfy the usual free
(from gravity) field equations, since the vielbein has been eliminated (the determi-
nant by redefinition, the rest by choice of general coordinate gauge). Similar remarks
apply to (8.1.28), but it’s not scale invariant. To isolate the scale noninvariance of
that action, rather than make the above scale transformation, we make a nonlocal
redefinition of χ in (8.1.28a) which reduces to the above type of scale transformation
in the conformal gauge ea m = δ a m :
χ → χ − (w − 12 ) R . (8.1.31)
In the conformal gauge, R = − 12 2 ln e. (Remember: χ is like the logarithm of a
tensor.) Under this redefinition, the action becomes
S→ d2 x e−1 1
2 χ − 12 (w − 12 )2 R
2χ R . (8.1.32)
(Note that χ now satisfies the usual scalar field equation.) The redefined field is now
scale invariant, and the scale noninvariance can now be attributed to the second term,
8.2. OSp(1,1|2) 157

which is the same kind of term responsible for the conformal (Virasoro) anomalies at
the quantum level (i.e., the 1-loop contribution to the 2D field theory in a background
gravitational field). In fact, the conformally invariant action (4.1.39), with a factor
proportional to 1/(D − 2), is the dimensionally regularized expression responsible for
the anomaly: Although (4.1.39) is conformally invariant in arbitrary D, subtraction
of the divergent (i.e., with a coefficient 1/(D − 2)) R term, which is conformally
invariant only in D = 2 (as follows from considering the D → 2 limit of (4.1.39)
without multiplying by 1/(D − 2)), leaves a renormalized (finite) action which, in the
limit D → 2, is just the second term of (8.1.32). Thus, the second term in (8.1.32)
contributes classically to the anomaly of (8.1.23b), the remaining contribution being
the usual quantum contribution of the scalar. (On the other hand, in the fermionic
theory from which (8.1.28) can be derived by quantum mechanical bosonization, all
of the anomaly is quantum mechanical.)

D < 26 can also be quantized (at least at the tree level), but there is an anomaly
in 2D local scale invariance which causes det(g mn ) to reappear at the quantum level
[8.9] (or, in the light-cone formalism, an extra “longitudinal” Lorentz component of
X [1.3,4]); however, there are complications at the one-loop level which have not yet
been resolved.

Presently the covariant formulation of string interactions is understood only
within the IGL(1) formalism (although in principle it’s straightforward to obtain
the OSp(1,1|2) formalism by eliminating the auxiliary fields, as in sect. 4.2). These
interactions will be discussed in sect. 12.2.

8.2. OSp(1,1|2)

We next use the light-cone Poincaré algebra of the string, obtained in sect. 7.1, to
derive the OSp(1,1|2) formulation as in sect. 3.4, which can be used to find the gauge-
invariant action. We then relate this to the first-quantized IGL(1) of the previous
section by the methods of sect. 4.2. This OSp(1,1|2) formalism can also be derived
from first quantization simply by treating the zero-mode of g as g for the particle
(sect. 5.1), and the other modes of the metric as in the conformal gauge (sect. 6.2).

All operators come directly from the light-cone expressions of sect. 7.1, using
the dimensional extension of sect. 2.6, as described in the general case in sect. 3.4.
In particular, to evaluate the action and its gauge invariance in the form (4.1.6),
we’ll need the expressions for M αβ , M 2 , and Qα = M α a pa + M αm M given by using

i = (a, α) in (7.1.14,15). Thus,
M αβ = −i X̂ α P β = a† n(α anβ) ,
M αa = −i X̂ α P a = a† n[α ana] ,

M = 2
(P i 2 − 2) = 2 na† n i ani − 1 = 2(N − 1) ,

M αm M = −i 12 X̂ α P j 2 = a† nα ∆n − ∆† n anα ,
 n−1 ∞

1 †
∆n = i √ 2 m(n − m)am an−m,i − i
m(n + m)a m
an+m,i , (8.2.1)
n m=1 m=1
where again the i summation is over both a and α, representing modes coming from
both the physical X a (σ) (with xa identified as the usual spacetime coordinate) and
the ghost modes X α (σ) (with xα the ghost coordinates of sect. 3.4), as in the mode
expansion (7.1.7).
To understand the relation of the first-quantized BRST quantization [4.4,5] to
that derived from the light cone (and from the OSp(1,1|2)), we show the Sp(2) sym-
metry of the ghost coordinates. We first combine all the ghost oscillators into an
“isospinor” [4.1]:  
1 δ
Cα = C  ,  . (8.2.2)
∂/∂σ δC
This isospinor directly corresponds (except for lack of zero modes) to X̂ α of the
OSp(1,1|2) formalism from the light cone: We identify

X̂ α = (xα + pα σ) + C α , (8.2.3)

and we can thus directly construct objects which are manifestly covariant under the
Sp(2) of M αβ . The periodic inverse derivative in (8.2.2) is defined in terms of the
saw-tooth function
1 1
f (σ) = dσ  12
(σ − σ ) − (σ − σ  ) f (σ  ) ,

∂/∂σ π
(σ) = ±1 for ±σ > 0. The product of the derivative with this inverse
derivative, in either order, is simply the projection operator which subtracts out
the zero mode. (For example, C + is just C minus its zero mode.) Along with P a ,
this completes the identification of the nonzero-modes of the two formalisms. We can
then rewrite the other relevant operators (8.1.10,12) in terms of C α :

2 2 dσ  2
p +M = (P a + C α C α  − 2) ,

8.2. OSp(1,1|2) 159

Qα = −i 12 C α (P a 2 + C α C α  − 2) ,

M αβ = −i 12 C (α C β)  . (8.2.5)

Again, all definitions include normal ordering. This first-quantized IGL(1) can then
be seen to agree with that derived from OSp(1,1|2) in (8.2.1) by expanding in zero-

For the closed string, the OSp(1,1|2) algebra is extended to an IOSp(1,1|2) algebra
following the construction of (7.1.17): As for the open-string case, the D-dimensional
indices of the light-cone formalism are extended to (D+4)-dimensional indices, but
just the values A = (±, α) are kept for the BRST algebra. To obtain the analog of
the IGL(1) formalism, we perform the transformation (3.4.3a) for both left and right-
handed modes. The extension of IGL(1) to GL(1|1) analogous to that of OSp(1,1|2)
to IOSp(1,1|2) uses the subalgebras (Q, J 3 , p− , pc̃ = ∂/∂c) of the IOSp(1,1|2)’s of each
set of open-string operators. After dropping the terms containing ∂/∂c̃, x− drops out,
and we can set p+ = 1 to obtain:

Q → − ic 12 (pa 2 + M 2 ) + M + i + Q+ ,

J3 → c + M3 ,
p− → − 12 (pa 2 + M 2 ) ,

pc̃ → . (8.2.6)

These generators have the same algebra as N=2 supersymmetry in one dimension,
with Q and pc̃ corresponding to the two supersymmetry generators (actually the
complex combination and its complex conjugate), J 3 to the O(2) generator which
scales them, and p− the 1D momentum. The closed-string algebra GL(1|1) is then
constructed in analogy to the IOSp(1,1|2), taking sums for the J’s and differences for
the p’s.

The application of this algebra to obtaining the gauge-invariant action will be
discussed in chapt. 11.

8.3. Lorentz gauge
We will next consider the OSp(1,1|2) algebra which corresponds to first-quantization
in the Lorentz gauge [3.7], as obtained by the methods of sect. 3.3 when applied to
the Virasoro algebra of constraints.
From (3.3.2), for the OSp(1,1|2) algebra we have
dσ δ δ δ
Q = −i
C α ( 12 P 2 − 1) + 12 C (α C β) i β − Bi + 12 (C α B  − C α B)i
2π δC δC α δB

+ 14 C β (C β  C α ) i . (8.3.1)

B is conjugate to the time-components of the gauge field, which in this case means
the components g00 and g01 of the unit-determinant part of the world-sheet metric
(see chapt. 6). This expression can be simplified by the unitary transformation Qα →
UQα U −1 with 
ln U = − 12 ( 12 C α C α ) . (8.3.2)
We then have the OSp(1,1|2) (from (3.3.7)):
δ δ δ
J −α = Cα −i 12 P 2 +i+C β
+ Bβ
−B ,
δC δB δC α
δ δ
J αβ = C (α β) , J +α = 2 Cα ,
δC δB

δ δ
J −+ = 2B + Cα α . (8.3.3)
δB δC

A gauge-fixed kinetic operator for string field theory which is invariant under the
full OSp(1,1|2) can be derived,
δ 1 2 δ δ
K= 1
J − , J −α , i = 2P − 1 + C α i α
+ Bi = 12 (p2 + M 2 ) ,
δB δC δB

as the zero-mode of the generators G(σ) from (3.3.10):
G(σ) = − 12 J− α
, J −α ,
δ δ δ δ
= −i( 12 P 2 − 1) + C α
+ Cα α +B + B
. (8.3.5)
δC δC δB δB
The analog in the usual OSp(1,1|2) formalism is
1 2
2 {J − , [J −α , p+ ]}
= pA pA ≡ 2p+ p− + pα pα = 2 − M 2 . (8.3.6)
8.3. Lorentz gauge 161

This differs from the usual light-cone gauge-fixed hamiltonian p− , which is not OSp(1,1|2)
invariant. Unlike the ordinary BRST case (but like the light-cone formalism), this
operator is invertible, since it’s of the standard form K = 12 p2 + · · ·. This was made
possible by the appearance of an even number (2) of anticommuting zero-modes. In
ordinary BRST (sect. 4.4), the kinetic operator is fermionic: c(2 − M 2 ) − 2 ∂c

M + is
not invertible because M + is not invertible.
As usual, the propagator can be converted into a form amenable to first-quantized
path-integral techniques by first introducing the Schwinger proper-time parameter:
= dτ e−τ K , (8.3.7)
K 0

where τ is identified with the (Wick-rotated) world-sheet time. At the free level, the
analysis of this propagator corresponds to solving the Schrödinger equation or, in
the Heisenberg picture (or classical mechanics of the string), to solving for the time
dependence of the coordinates which follow from treating K as the hamiltonian:

[K, Z] = iZ  ,
Z − [K, Z] = 0 → Z = Z(σ + iτ ) (8.3.8)
for Z = P, C, B, δ/δC, δ/δB. Thus in the mode expansion Z = z 0 + 1 (z n † e−in(σ+iτ ) +
z n ein(σ+iτ ) ) the positive-energy z n † ’s are creation operators while the negative-energy
z n ’s are annihilation operators. (Remember active vs. passive transformations: In the
Schrödinger picture coordinates are constant while states have time dependence e−tH ;
in the Heisenberg picture states are constant while coordinates have time dependence
etH ( )e−tH .)
When doing string field theory, in order to define real string fields we identify
complex conjugation of the fields as usual with proper-time reversal in the mechanics,
which, in order to preserve handedness in the world sheet, means reversing σ as well
as τ . As a result, all reparametrization-covariant variables with an even number of
2D vector indices are interpreted as string-field coordinates, while those with an odd
number are momenta. (See sect. 8.1.) This means that X is a coordinate, while B and
C are momenta. Therefore, we should define the string field as Φ[X(σ), G(σ), F α (σ)],
where B = iδ/δG and C α = iδ/δF α . This field is real under a combined complex
conjugation and “twist” (σ → −σ), and Qα is odd in the number of functional plus
σ derivatives. (Note that the corresponding replacement of B with G and C with F
would not be required if the G i ’s had been associated with a Yang-Mills symmetry
rather than general coordinate transformations, since in that case B and C carry no
vector indices.)

This OSp(1,1|2) algebra can also be derived from the classical mechanics action.
 dσ m
The 2D general coordinate transformations (6.1.2) generated by δ =
2π m (σ)

determine the BRST transformations by (3.3.2):

Qα X = C mα ∂ m X ,

Qα gmn = ∂ p (C pα gmn ) − gp(m ∂ p C n)α ,

Qα C mβ = 12 C n(α ∂ n C mβ) − C αβ B m ,

Qα B m = 12 (C nα ∂ n B m − B n ∂ n C mα )

− 8 2C nβ (∂ n C pα )∂ p C m β + C nβ C p β ∂ n ∂ p C mα .


We then redefine
B m = B m − 12 C nα ∂ n C m α →

Qα C mβ = C nα ∂ n C mβ − C αβ B ,

Qα B m = C nα ∂ n B
m . (8.3.10)

The rest of the OSp(1,1|2) follows from (3.3.7):

J +α (X, gmn , C mβ ) = 0 ,  m = 2C m
J +α B ;

 m) = 0 ,
J αβ (X, gmn , B J αβ C mγ = δ (α γ C m β) ;

J −+ (X, gmn ) = 0 ,  m = 2B
J −+ B m , J −+ C mα = C mα . (8.3.11)

An ISp(2)-invariant gauge-fixing term is (dropping boundary terms)

L1 = Qα 2 21 η pq gpq = −η pq B p ∂ m gqm + 12 gmn (∂ m C pα )(∂ n C q α ) , (8.3.12)

where η is the flat world-sheet metric. This expression is the analog of the gauge-
 gives the
fixing term Q2 12 A2 for Lorentz gauges in Yang-Mills [3.6,12]. Variation of B
condition for harmonic coordinates. The ghosts have the same form of lagrangian
as X, but not the same boundary conditions: At the boundary, any variable with
an even number of 2D vector indices with the value 1 has its σ-derivative vanish,
while any variable with an odd number vanishes itself. These are the only boundary
conditions consistent with Poincaré and BRST invariance. They are preserved by the
redefinitions below.
8.3. Lorentz gauge 163

Rather than this Landau-type harmonic gauge, we can also define more general
Lorentz-type gauges, such as Fermi-Feynman, by adding to (8.3.12) a term propor-
tional to Q2 12 η mn C mα C n α = − 12 η mnB m B n + · · ·. We will not consider such terms
further here.
Although the hamiltonian quantum mechanical form of Qα (8.3.3) also follows
from (3.3.2) (with the functional derivatives now with respect to functions of just σ
instead of both σ and τ ), the relation to the lagrangian form follows only after some
redefinitions, which we now derive. The hamiltonian form that follows directly from
(8.3.9,10) can be obtained by applying the Noether procedure to L = L0 + L1 : The
BRST current is

J mα = (gnp C mα − gm(n C p)α ) 12 (∂ n X) · (∂ p X) + (∂ n C qβ )(∂ p C qβ )

− B n ∂ p (gn[m C p]α − gmp C nα ) , (8.3.13)

where (2D) vector indices have been raised and lowered with the flat metric. Canon-
ically quantizing, with
1 0 δ 1  δ 1 δ

P =i , 
B m = −i 0m , 
Πmα = i mα , (8.3.14a)
α δX α δg α δC
we apply
(P 0 + g01 X  ) ,
C mα = −
(Πmα + g01 C mα ) , (8.3.14b)
g00 g00
to the first term in (8.3.13) and

∂ m gmn = 0 , g0m ∂ m C nα = −Πnα , (8.3.14c)

to the second to obtain
J 0α
= − 00 C 0α 1
2 (P 02 + X 2 ) + (Πmβ Πmβ + C mβ  C mβ  )
+ C 1α 
− 00 C 0α (X  · P 0 + C mβ  Πmβ ) − B mα
+ (gm[0 C 1]α) ,
m Π

where Qα ∼ dσ J 0α .
By comparison with (8.3.3), an obvious simplification is to absorb the g factors
into the C’s in the first two terms. This is equivalent to

C mα → δ 1 m C 1α − g0m C 0α , (8.3.16)

 This puts
and the corresponding redefinitions (unitary transformation) of Π and B.
g-dependence into the BΠ
1 0 g01
Πm B m → Π0 − 00 B + Π1 B 1 − 00 B 0 + · · · , (8.3.17)
g g
unlike (8.3.3), so we remove it by the g redefinition

g01 = −g 1 , g00 = − 1 + 2g 0 + (g 1 )2 ;

1 δ

Bm = i m . (8.3.18)
α δg
These redefinitions give

J 0α = C 0α 12 (P 02 + X 2 ) + C 1α X  · P 0
+ C 0α (C 0β Π1β ) + C 1α C mβ  Πmβ + (C 0β Π0β )
− C 0α (B 1 + g 1 B 0 ) + C 1α −B 0 + g m B m + (g 0 B 0 )

1 mβ 
+ C 0α 2C C mβ  + (g m C 0β ) C m β  − Πm α B m . (8.3.19)

The quadratic terms CB  don’t appear in (8.3.3), and can be removed by the unitary

−1 1
Q α
= UQ U α
, ln U = −i  C 0α C 1 α  , (8.3.20)

J 0α = C 0α 12 (P 02 + X 2 ) + C 1α X  · P 0
+ C 0α (C 0β Π1β ) + C 1α C mβ Πmβ + (C 0β Π0β )
− C 0α (g 1 B 0 ) + C 1α g m B m + (g 0 B 0 )

+ C 0α (g m C 0β ) C m β  − Πm α B m . (8.3.21)

Finally, the remaining terms can be fixed by the transformation

ln U = i  C 0α (g 1 C 0 α  + g 0 C 1 α  ) (8.3.22)
to get (8.3.3), after extending σ to [−π, π] by making the definitions
P̂ = √ 
(P 0 + X  ) ,

1 δ 1
Cα = √ (C 0α + C 1α ) , i = √  (Π0α + Π1α ) ,
2α δC α

8.3. Lorentz gauge 165

1 δ 1
G= √ 
(g 0 + g 1 ) , i = √  (B 0 + B 1 ) , (8.3.23)
2α δG 2α
where the previous coordinates, defined on [0, π], have been extended to [−π, π] as
Z(−σ) = ±Z(σ) , (8.3.24)
with “+” if Z has an even number of vector indices with the value 1, and “−” for an
odd number, in accordance with the boundary conditions, so that the new coordinates
will be periodic in σ with period 2π.
To describe the closed string with the world-sheet metric, we again use 2 sets of
open-string operators, as in (7.1.17), with each set of open-string operators given as
in (8.3.3,4), and the translations are now, in terms of the zero-modes b and c of B
and C,

p+ = −2i,
1 ∂
pα = i ,
p+ ∂cα

p− = − K + pα 2 . (8.3.25)
The OSp(1,1|2) subgroup of the resulting IOSp(1,1|2), after the use of the constraints
∆p = 0, reduces to what would be obtained from applying the general result (3.3.2)
to closed string mechanics. ((3.2.6a), without the B∂/∂ C̃ term, gives the usual BRST
operator when applied to closed-string mechanics, which is the same as the sum of
two open-string ones with the two sets of physical zero-modes identified.)
We now put the OSp(1,1|2) generators in a form analogous to those derived
from the light cone [3.13]. Let’s first consider the open string. Separating out the
dependence on the zero-modes g and f α in the OSp(1,1|2) operators,
J αβ = f (α β) + M αβ ,
∂ ∂ 
J −+ = −2g − fα α + M −+ ,
∂g ∂f
J +α = −2g α + M +α ,
∂ ∂ 
J −α = f α − igBα + K α + iC α β f β + M −α , (8.3.26)
∂g ∂f
 are the parts of the J’s not containing these zero-modes, K is given
where the M’s
in (8.3.4), and

Bα = BC α  ,

C αβ = C α C β  = C βα . (8.3.27)

 , B , C , and K; the
From (3.3.8) and (8.3.26) we find the commutators of M AB α αβ

nonvanishing ones are:

[M  ] = −C 
αβ γδ (γ(α M β)δ) ,
[M  ] = −C M  ,
αβ ±γ γ(α ±β)

{M  } = −C M  
−α +β αβ −+ − M αβ ,
[M  ] = ∓M
−+ ±α ±α

{M  } = 2iKC ,
−α −β αβ

 , B ] = −C B  , C ] = −C
[M αβ γ γ(α β) , [M αβ γδ (γ(α C β)δ) ,
 , B ] = 3B
[M ,  , C ] = 2C
[M ,
−+ α α −+ αβ αβ

 , B } = 2C
{M  , C ] = −C
+α β αβ , [M −α βγ α(β B γ) . (8.3.28)

(To show {M −[α , B β] } = 0 requires explicit use of (8.3.3), but it won’t be needed


We then make the redefinition (see (3.6.8))

g = 12 h2 (8.3.29)

as for the particle. We next redefine the zero-modes as in (3.5.2) by first making the
unitary transformation
1 ∂ 
ln U = (ln h) 2 , fα − M −+ (8.3.30a)
∂f α

to redefine h and then

ln U = −f α M (8.3.30b)

to redefine cα . The net result is that the transformed operators are
∂ ∂ ∂ 
J −+ = − h , J +α = −h , J αβ = f (α +M αβ ,
∂h ∂f α ∂f β)
∂ 1 ∂  βf + Q

J −α = fα + (K + f 2 ) − M α β α , (8.3.31a)
∂h h ∂f α
Q  1 
α −α − i 2 B α + K M +α . (8.3.31b)
8.3. Lorentz gauge 167

We also have
p+ = h , p− = − (K + f 2 ) .
pα = −f α , (8.3.32)
These expressions have the canonical form (3.4.2a), with the identification

M ,  ↔Q
Q . (8.3.33)
αβ αβ α α

From (8.3.28) we then find

{Q  } = −2K M
 , (8.3.34)
α β αβ

consistent with the identification (8.3.33). Thus the IOSp(1,1|2) algebra (8.3.31,32)
takes the canonical form of chapt. 3. This also allows the closed string formalism to
be constructed.

For expanding the fields, it’s more convenient to expand the coordinate in real
oscillators (preserving the reality condition of the string field) as
∞ √

P̂ = p + n(−ian † e−inσ + ian einσ ) ,

∞ √

G=g+ n(g n † e−inσ + g n einσ ) ,

B = b+ √ (−ibn † e−inσ + ibn einσ ) ,
1 n
∞ √

Fα = fα + n(f α n † e−inσ + f α n einσ ) ,

C α = cα + √ (−icα n † e−inσ + icα n einσ ) . (8.3.35)
1 n
(With our conventions, always z n α† ≡ (z n α )† , and thus z nα † = −(z nα )† .) The com-
mutation relations are then

[am , an † ] = [bm , g n † ] = [g m , bn † ] = δ mn ,

{cmα , f nβ † } = {f mα , cnβ † } = δ mn C αβ , (8.3.36)

We then have

Q = (On † cnα − cnα † On ) − (bn † f nα − f nα † bn ) ,

1  1
On = √ L n + 2 bn − 2g n K ,

 = −12 − 1 +
K=L n(an † · an + bn † g n + g n † bn + cn α† f nα + f n α† cnα )
0 2

= − 12 2 −1+N ,

 = −

L n n∂ · an

+ m(n + m)(am † · an+m + bm † g n+m + cm α† f n+m,α

+ g m † bn+m + f m α† cn+m,α )

+ m(n − m)(− 12 am · an−m + bm g n−m + cm α f n−m,α ) . (8.3.37)

The Lorentz-gauge OSp(1,1|2) algebra can also be derived by the method of sect.
3.6. In the GL(1) case, we get the same algebra as above, while in the U(1) case we get
a different result using the same coordinates, suggesting a similar first-quantization
with the world-sheet metric. For the GL(1) case we define new coordinates at the
GL(2|2) stage of reduction:

X A = x̃A + CA , P A = p̃A + C A , (8.3.38)

where C is the generalization of C α of (8.2.3) to arbitrary index. (Remember that in
GL(2|2) a lower primed index can be converted into an upper unprimed index.) P A
is then canonically conjugate to X A . We also have relations between the coordinates
such as
P A = P A , P A = p̌A + X A ,
A = p̌A σ + X A
X , (8.3.39)
where P A = (P A , P A ), etc. However, X
 can be expressed in terms of P only

with an inverse σ-derivative. As a result, when reexpressed in terms of these new
coordinates, of all the IOSp(2,2|4) generators only the IGL(2|2) ones have useful
expressions. Of the rest, J˜AB is nonlocal in σ, while J˜AB and p̃A have explicit separate
terms containing xA and pA . Explicitly, the local generators are pA , p̌A , and
J˜A B = i(−1)AB x̌B p̌A − i X AP B , J˜AB = −ix[A p̌B) − i X A X B . (8.3.40)

In these expressions we also use the light-cone constraint and gauge condition as
translated into the new coordinates:
1  2 + + α

P− = − P a − 2 + 2(p̌ + X )P + + 2(p̌ α
+ X )P α , X− = 0 .
8.3. Lorentz gauge 169

After following the procedure of sect. 3.6 (or simply comparing expressions to deter-
mine the M ij ), the final OSp(1, 1|2) generators (3.6.7c) are (dropping the primes on
the J’s)
J +α = 2i X αP + , J −+ = −i (2X − P + + X α P α ) , J αβ = −i X (α P β) ,

J −α = −i [X − P α + X α ( 12 P a 2 − 1 + X  − P + + X β P β )] . (8.3.42)

This is just (8.3.3), with the identification P + = G and X α = C α .

In the U(1) case there is no such redefinition possible (which gives expressions
local in σ). The generators are
J +α = ixα p+ + i X̂ α 
P , J −+ = −ix− p+ − i X̂ − P + ,

J αβ = −ix(α pβ) − i X̂ α P β −i X̂ α P β  ,

J −α = − ix− pα + ixα P − + i X̂ α P −
−i pa X̂ α P a −p β
X̂ α P β −i X̂ − P α ,

−2p+ P − = P a 2 − 2 + P α P α + P α P α + 2P + P − , (8.3.43)

where xA = pA = 0. After performing the unitary transformations (3.6.13), they
J +α = ixα p+ , J −+ = −ix− p+ , J αβ = −ix(α pβ) − i X̂ α P β − i X̂ α P β  ,

J −α = − ix− pα + ixα P − +i X̂ α P −
1 a
−i p X̂ α P a −p β
X̂ α P β + X̂ α 
P β
+ X̂ − 
P + 1
X̂ + 
pa − 2p+ P . (8.3.44)
α 2 α −

We can still interpret the new coordinates as the world-sheet metric, but different
redefinitions would be necessary to obtain them from those in the mechanics action,
and the gauge choice will probably differ with respect to the zero-modes.

Introducing extra coordinates has also been considered in [8.10], but in a way
equivalent to using bosonized ghosts, and requiring imposing D = 26 by hand instead

of deriving it from unitarity. Adding 4+4 extra dimensions to describe bosonic strings
with enlarged BRST algebras OSp(1, 1|4) and OSp(2, 2|4) has been considered by
Aoyama [8.11].

The use of such extra coordinates may also prove useful in the study of loop dia-
grams: In particular, harmonic-type gauges can be well-defined globally on the world
sheet (unlike conformal gauges), and consequently certain parameters (the world-
sheet generalization of proper-time parameters for the particle, (5.1.13)) appear au-
tomatically [8.12]. This suggests that such coordinates may be useful for closed string
field theory (or superstring field theory, sect. 7.3).

The gauge-invariant action, its component analysis, and its comparison with that
obtained from the other OSp(1,1|2) will be made in sect. 11.2.


(1) Prove that the operators (8.1.9) satisfy commutation relations like (7.1.7a). Prove
that they are conformally invariant.

(2) Derive (8.1.20). Verify that the usual fermionic anticommutation relations for
(8.1.18) then follow from (8.1.19). Derive (8.1.22).

(3) Derive (8.1.15). Derive (8.1.23bc). Prove Q2 = 0.

(4) Derive (8.1.23ac) from the energy-momentum tensors of (8.1.27,28ac).

(5) Find the commutation relations of 12 D̂dD̂ of (7.2.2), generalizing (8.1.2). Find
the BRST operator. Derive the gauge-fixed Virasoro operators, and show the
conformal weights of Ψ̂ is 1/2, and of its ghosts are −1/2 and +3/2. Use (8.1.23)
to show the anomaly cancels for D = 10.

(6) Find an alternate first-order form of (8.1.28b) by rewriting the last term in terms
of F and ω, and show how (8.1.28ac) follow.

(7) Show from the explicit expression for ω a that (8.1.27) can be made vielbein-
independent by field redefinition in the conformal gauge.

(8) Explicitly prove the equivalence of the IGL(1)’s derived in sect. 8.1 and from the
light cone, using the analysis of sect. 8.2.

(9) Derive (8.3.1,3). Derive (8.3.9).

(10) Derive (8.3.28).
9.1. External fields 171


9.1. External fields

One way to derive Feynman graphs is by considering a propagator in an external


For example, for a scalar particle in an external scalar field,

L = 12 ẋ2 − m2 − φ(x) → (9.1.1a)

[p2 + m2 + φ(x)]ψ(x) = 0 → (9.1.1b)
1 1 1 1
propagator = − 2 φ 2 +··· (9.1.1c)
p2 + m2 + φ(x) p2 +m2 p + m p + m2

= +  +   +···

For the (open, bosonic) string, it’s useful to use the “one-handed” version of X(σ)
(as in (7.1.7b)) so that X̂ and P can be treated on an (almost) equal footing, so we
will switch to that notation hand and foot. Then the generalization of (9.1.1b) (again
jumping directly to the first-quantized form for convenience) is [7.5]
χ[ 12 P 2 (σ) − 1 + V(σ)]ψ =0 . (9.1.2)

(The OSp(1,1|2) algebra can be similarly generalized.) However, for consistency of
these equations of motion, they must satisfy the same algebra as 12 P 2 − 1 ((8.1.2)).
(In general, this is only true including ghost contributions, which we will ignore for
172 9. GRAPHS

the examples considered here.) Expanding this condition order-by-order in V, we get
the new relations

[ 12 P 2 (σ 1 ) − 1, V(σ 2 )] = 2πiδ  (σ 2 − σ 1 )V(σ 1 )

= 2πi [δ(σ 2 − σ 1 )V  (σ 2 ) + δ  (σ 2 − σ 1 )V(σ 2 )] , (9.1.3a)

[V(σ 1 ), V(σ 2 )] = 0 . (9.1.3b)

The first condition gives the conformal transformation properties of V (it transforms
covariantly with conformal weight 1, like P ), and the second condition is one of
“locality”. A simple example of such a vertex is a photon field coupled to one end of
the string:
V(σ) = 2πδ(σ)A (X(0)) · P , (9.1.4)

where X(0) = X̂(0). Graphs are now given by expanding the propagator as the
inverse of the hamiltonian
dσ 1  2
H= ( P − 1 + V) = L0 + dσ V ≡ L0 + V . (9.1.5)
2π 2

More general vertices can be found when normal ordering is carefully taken into
account [1.3,4], and one finds that (9.1.3a) can be satisfied when the external field is
on shell. For example, consider the scalar vertex

V(σ) = −2πδ(σ)φ (X(0)) . (9.1.6)

Classically, scalar fields have the wrong conformal weight (zero):
1 1 2

P (σ 1 ) − 1, φ X̂(σ 2 ) = iδ(σ 2 − σ 1 ) φ X̂(σ 2 ) ; (9.1.7a)
2π 2 ∂σ 2

but quantum mechanically they have weight “− 12 2 ”:

1 1 2

2 P (σ 1 ) − 1, φ X̂(σ 2 ) = iδ(σ 2 − σ 1 ) φ X̂(σ 2 )
2π ∂σ 2

+ −2 1
iδ  (σ 2 − σ 1 )φ X̂(σ 2 ) .
∂ X̂ 2 (σ 2 )

Therefore φ X̂(σ 2 ) , and thus V of (9.1.6), satisfies (9.1.3a) if φ is the on-shell ground
state (tachyon): − 12 2 φ = 1 · φ. (Similar remarks apply quantum mechanically for
the masslessness of the photon in the vertex (9.1.4).)
9.1. External fields 173

As an example of an S-matrix calculation, consider a string in an external plane-
wave tachyon field, where the initial and final states of the string are also tachyons:

φ(x) = e−ik·x → : e−ik·X(0) : (9.1.8)

We then find

k1 kN
k2 k N−1

= g N−2 k N | V (k N−1 ) · · · ∆(p)V (k 3 )∆(p)V (k 2 ) |k 1

= g N−2 0 V (k N−1 ) · · · ∆(k 3 + k 4 + · · · + k N )V (k 2 ) 0 ,

V (k) = : e −ik·X(0)
: = V (k)e−ik·x , X(0) = √ (an † + an ) ,
1 n

∆(p) = 1 2 , N= nan † · an , (9.1.9)
2p + (N − 1) 1
where g is the coupling constant, and we have pulled the x pieces out of the X’s and
pushed them to the right, causing shifts in the arguments of the ∆’s (which were
originally p, the momentum operator conjugate to x, not to be confused with the
constants k i ). We use Schwinger-like parametrizations (5.1.13) for the propagators:
 ∞ 1  1
1 −t[ 2 p2 +(N −1)] dx 12 p2 +(N −1)
1 2 = dt e = x (x = e−t ) , (9.1.10)
2 p + (N − 1) 0 0 x

where we use ti for ∆(k i + · · · + k N ), as the difference in proper time between V (k i−1 )
and V (k i ). Plugging (9.1.10) into (9.1.9), the amplitude is
N−1  '    (
" 1 dxi 21 (ki +···+kN )2 −1 "  na† ·a −ik 2 · √n a† 
g N−2
xi 0 · · · x3 e 0 . (9.1.11)
i=3 0 xi n

To evaluate matrix-elements of harmonic oscillators it’s generally convenient to use
coherent states:

≡ eza |0

   † †
a |z
= z |z
, a† z = z , ez a |z
= |z + z 
, xa a |z
= |xz

 z̄z  d2 z −|z|2
= e , 1= e |z
z| ,
174 9. GRAPHS

d2 z −|z|2
e z| O |z
tr(O) = (9.1.12)
Using (9.1.12) and the identity ∞
1 e
−cxn /n
= (1 − x)c , (9.1.11) becomes
   ki ·kj
" dxi 12 (ki +···+kN)2 −1 " "
g N−2 xi 1 − xk  . (9.1.13)
xi 2≤i<j≤N−1 k=i+1

We next make the change of variables

τi = tj , τ2 = 0 , τN = ∞ → 0 = τ2 ≤ τ3 ≤ · · · ≤ τN = ∞ ,
zi = xj , z2 = 1 , zN = 0 → 1 = z 2 ≥ z 3 ≥ · · · ≥ z N = 0 , (9.1.14b)

z i = e−τ i , (9.1.14c)

where τ i is the absolute proper time of the corresponding vertex. Using the mass-shell
condition k i 2 = 2, the final result is then [9.1]
" "
g N−2
dz i (z i − z j )ki ·kj . (9.1.15)
i=3 2≤i<j≤N

The simplest case is the 4-point function (N= 4) [9.2]
 1 1 1
g 2
dz z − 2 s−2 (1 − z)− 2 t−2 ( s = −(k 1 + k 2 )2 , t = −(k 2 + k 3 )2 )
Γ(− 12 s − 1)Γ(− 12 t − 1)
= g 2 B(− 12 s − 1, − 12 t − 1) = g 2 (9.1.16a)
Γ(− 12 s − 12 t − 2)

( t + 1 + j)( 1 t + j) · · · ( 1 t + 1)
2 2 2 1
=g (9.1.16b)
j=0 j! j− ( 12 s + 1)
 
= in s→−∞
lim  g 2 Γ(− 12 t − 1)(− 12 s − 1) 2 t+1 . (9.1.16c)
t f ixed

(9.1.16b) shows that the amplitude can be expressed as a sum of poles in the s channel
with squared masses 2(j − 1), with maximum spin j (represented by the coefficient
with leading term tj ). Since the amplitude is symmetric in s and t, it can also be
expressed as a sum of poles in the t channel, and thus summing over poles in one
channel generates poles in the other. (It’s not necessary to sum over both.) This
9.1. External fields 175

property is called “duality”. (9.1.16c) shows that the high-energy behavior goes like
s 2 t+1 instead of the usual fixed-power behavior sj due to the exchange of a spin j
particle, which can be interpreted as the exchange of a particle with effective spin
j = 12 t + 1. This property is known as “Regge behavior”, and j(t) = 12 t + 1 is called
the “leading Regge trajectory”, which not only describes the high-energy behavior
but also the (highest) spin at any given mass level (the mass levels being given by
integral values of j(t)).

Instead of using operators to evaluate propagators in the presence of external
fields, we can also use the other approach to quantum mechanics, the Feynman path
integral formalism. In particular, for the calculation of purely tachyonic amplitudes
considered above, we evaluate (9.1.9) directly in terms of V (rather than V ), after
using (9.1.10):
 ∞ '   (
 1 2 +N −1) 
g N−2 dt3 · · · dtN−1 k N V (k N−1 ) · · · e−t3 ( 2 p V (k 2 ) k 1 . (9.1.17)

Using (9.1.14a), we can rewrite this as
" τ i+1
g N−2
dτ i k N | V (k N−1 , τ N−1 ) · · · V (k 3 , τ 3 )V (k 2 , τ 2 ) |k 1
, (9.1.18a)
i=3 τ i−1

1 2 +N −1) 1 2 +N −1)
V (k, τ ) = : e−ik·X(0,τ ) : , X(0, τ ) = eτ ( 2 p X(0)e−τ ( 2 p , (9.1.18b)

is the vertex which has been (proper-)time-translated from 0 to τ . (Remember that in
the Heisenberg picture operators have time dependence O(t) = etH O(0)e−tH , whereas
in the Schrödinger picture states have time dependence |ψ(t)
= e−tH |ψ(0)
, so that
time-dependent matrix elements are the same in either picture. This is equivalent to
the relation between first- and second-quantized operators.) External states can also
be represented in terms of vertices:

= lim V (k, τ )e−τ |0
, k| = lim 0| eτ V (k, τ ) . (9.1.19)
τ →−∞ τ →∞

The amplitude can then be represented as, using (9.1.14c),
" z i+1
g N−2
dz i lim (z 1 )2 0| V  (k N , z N ) · · · V  (k 1 , z 1 ) |0

z 1 →∞
, (9.1.20a)
i=3 z i−1 z N →0

V  (k, z) = − V (k, τ (z)) (9.1.20b)
176 9. GRAPHS

according to (8.1.6), since vertices have weight w = 1. The amplitude with this form of
the external lines can be evaluated by the same method as the previous calculation.
(In fact, it directly corresponds to the calculation with 2 extra external lines and
vanishing initial and final momenta.) However, being a vacuum matrix element, it
is of the same form as those for which path integrals are commonly used in field
theory. (Equivalently, it can also be evaluated by the operator methods commonly
used in field theory before path integral methods became more popular there.) More
details will be given in the following section, where such methods will be generalized
to arbitrary external states.
Coupling the superstring to external super-Yang-Mills is analogous to the bosonic
string and superparticle [2.6]: Covariantize D α → D α + δ(σ)Γα , P a → P a + δ(σ)Γa ,

Ωα → Ωα + δ(σ)W α . Assuming dσ A as kinetic operator (again ignoring ghosts),
the vertex becomes
V = W α D α + Γa P a − Γα Ωα (9.1.21)

evaluated at σ = 0. Solving the constraints (5.4.8) in a Wess-Zumino gauge, we find

W α ≈ λα ,

Γa ≈ Aa + 2γ a αβ Θα λβ ,

Γα ≈ γ a αβ Θβ Aa + 43 γ a αβ γ aγδ Θβ Θγ λδ , (9.1.22)

evaluated at σ = 0, where “≈” means dropping terms involving x-derivatives of the
physical fields Aa and λα . Plugging (7.3.5) and (9.1.22) into (9.1.21) gives
V ≈A a
P a +λ α
− γ a αβ P a Θβ − 16 iγ a αβ γ aγδ Θβ Θγ Θδ . (9.1.23)
Comparing with (7.3.6), we see that the vertices, in this approximation, are the same
as the integrands of the supersymmetry generators, evaluated at σ = 0. (In the case
of ordinary field theory, the vertices are just the supersymmetry generators pa and q α ,
to this order in θ.) Exact expressions can be obtained by expansion of the superfields
Γα , Γa , and W α in (9.1.21) to all orders in Θ [7.6]. In practice, superfield techniques
should be used even in the external field approach, so such explicit expansion (or even
(9.1.22) and (9.1.23)) is unnecessary. It’s interesting to note that, if we generalize D,
P , and Ω to gauge-covariant derivatives ∇α = D α +Γα , ∇a = D a +Γa , ∇α = Ωα +W α ,
with Γα , Γa , and W α now functions of σ, describing the vector multiplets of all
masses, then the fact that the only mode of the ∇’s missing is the zero-mode of Ωα 

( dσ Ωα = 0) directly corresponds to the fact that the only gauge-invariant mode
9.2. Trees 177

of the connections is the zero-mode of W α (the massless spinor, the massive spinors
being Stueckelberg fields).
The external field approach has also been used in the string mechanics lagrangian
method to derive field theory lagrangians (rather than just S-matrices) for the lower
mass levels (tachyons and massless particles) [9.3,1.16]. Since arbitrary external fields
contain arbitrary functions of the coordinates, the string mechanics lagrangian is no
longer free, and loop corrections give the field theory lagrangian including effective
terms corresponding to eliminating the higher-mass fields by their classical field equa-
tions. Thus, calculating all mechanics-loop corrections gives an effective field theory
lagrangian whose S-matrix elements are the tree graphs of the string field theory with
external lines corresponding to the lower mass levels. Such effective lagrangians are
useful for studying tree-level spontaneous breakdown due to these lower-mass fields
(vacua where these fields are nontrivial). Field-theory-loop corrections can be cal-
culated by considering more general topologies for the string (mechanics-loops are
summed for one given topology).

9.2. Trees
The external field approach is limited by the fact that it treats ordinary fields
individually instead of treating the string field as a whole. In order to treat general
string fields, a string graph can be treated as just a propagator with funny topology:
For example,


' $ 


& % 


can be considered as a propagator where the initial and final “one-string” states just
happen to be disconnected. The holes in the world sheet represent loops. When group
178 9. GRAPHS

theory indices are associated with the ends of the lines, the values of the indices are
required to be the same along the entire line, which corresponds to tracing in the ma-
trices associated with the string states. (The ends of the strings can be interpreted as
“quarks” which carry the “flavor” quantum numbers, bound by a string of “gluons”
which carry only “color” canceled by that of the quarks.) Such an approach is limited
to perturbation theory, since the string is necessarily gauge-fixed, and any one graph
has a fixed number of external lines and loops, i.e., a fixed topology. The advan-
tage of this approach to graphs is that they can be evaluated by first-quantization,
analogously to the free theory. (Even the second-quantized coupling constant can be
included in the first-quantized formalism by noting that the power of the coupling
constant which appears in a graph, up to wave function normalizations, is just the
Euler number, and then adding the corresponding curvature integral to the mechanics

We first consider the light-cone formalism. We Wick rotate the proper time τ →
iτ (see sects. 2.5-6), so now conformal transformations are arbitrary reparametriza-
tions of ρ = −τ + iσ (and the complex conjugate transformation on ρ̄) instead of
τ + σ (and of τ − σ independently: see (6.2.7)), since the metric is now dσ 2 = dρdρ̄.
There are three parts to the graph calculation: (1) expressing the S-matrix in terms
of the Green function for the 2D Laplace equation, (2) finding an explicit expression
for the Green function for the 2D surface for that particular graph, by conformally
transforming the ρ plane to the upper-half complex (z) plane where the Green func-
tion takes a simple form, and (3) finding the measure for the integration over the
positions of the interaction points.

The first step is easy, and can be done using functional integration [9.4,1.4] or
solving functional differential equations (the string analog of Feynman path integrals
and the Schrödinger equation, respectively). Since all but the zero-mode (the usual
spacetime coordinate) of the free string is described by an infinite set of harmonic
oscillators, the most convenient basis is the “number” basis, where the nonzero-modes
are represented in terms of creation and annihilation operators. The basic idea is then
to represent S-matrix elements as

A = ext|V
= ext| e∆ |0
, (9.2.1)

where |V
represents the interaction and ext| represents all the states (initial and
final) of the external strings, in the interaction picture. This is sort of a spacetime
symmetric version of the usual picture, where an initial state propagates into a final
9.2. Trees 179

state: Instead, the vacuum propagates into an “external” state. The exponential e∆
is then the analog of the S-matrix exp(−H IN T t), which propagates the vacuum at
time 0 to external states at time t. It thus converts annihilation operators on its left
(external, “out” states) into creation operators (for the “in” state, the vacuum, at
“time” x+ = 0). ∆ itself is then the “connected” S-matrix: In this first-quantized pic-
ture, which looks like a free 2D theory in a space with funny geometry, it corresponds
directly to the free propagator in this space. Since we work in the interaction picture,
we subtract out terms corresponding to propagation in an “ordinary” geometry.

In the former approach, the amplitude can be evaluated as the Feynman path

 
A=  dτ i  DX i (σ, τ )
" −i α1
P r (σ)·X r (σ,τ 1 )
· DP r (σ) Ψ[P r ]e 2π
d2 σ 1 2
− p−r τ 1r − α1 (Ẋ +X 2
·e 2π 2 , (9.2.2)

corresponding to the picture (e.g., for N= 5)



τ1 τ1

where the τ ’s have been Wick-rotated, τ 1r is the end of the rth string (to be taken
to ±∞), the p− τ 1 factor amputates external lines (converts from the Schrödinger
picture to the interaction picture), the factor in large brackets is the external-line wave
function Ψ[X] (or Ψ̄[X] for outgoing states) written as a Fourier transform, and the
constant corresponds to the usual normal-ordering constant in the free hamiltonian.
180 9. GRAPHS

For explicitness, we have written the integration over interaction points ( dτ ) for
the simple case of open-string tree graphs. Planar graphs always appear as rectangles
due to the string lengths being 2παp+ , which is conserved. The functional integral
(9.2.2) is gaussian, so, making the definition
J(σ, τ ) = iδ(τ − τ 1 ) √  P (σ) , (9.2.3)
we find
d2 σ 11 2 1 d2 σ d2 σ 
−  2 (∂X) + √  JX → − 12 J(σ)G(σ, σ  )J(σ  ) ,
2π α α 2π 2π
2  2  ∂
∂ G(σ, σ ) = 2πδ (σ − σ) , G(σ, σ  ) = f (σ) 2
dσJ∼ p=0 ,
σr σs 
G= (2 − δ m0 )(2 − δ n0 )G rs
mn cos m cos n emτ r /p+r +nτ s /p+s
p+r p+s
+ Gf ree + (zero-mode)2 terms
→ A= ( dτ i )V (τ ) Ψ e∆  0 , ∆= 1
4 Grs mn αr m · αs n , (9.2.4)
where G is the 2D Green function for the kinetic operator (laplacian) ∂ 2 /∂τ 2 +∂ 2 /∂σ 2
of that particular surface, and V (τ ) comes from det(G). We have used Neumann
boundary conditions (corresponding to (6.1.5)), where the ambiguity contained in
the arbitrary function f (necessary in general to allow a solution) is harmless because
of the conservation of the current J (i.e., the momentum p). The Gf ree term is
dropped in converting to the interaction picture: In functional notation (see (9.2.2)),
it produces the ground-state wave function. The (zero-mode)2 terms are due to
boundary conditions at ∞, and appear when the map to the upper-half plane is chosen
so that the end of one string goes to ∞, giving a divergence. They correspond to the
factor 1/z N in the similar map used for (9.1.20a). The factors (2 − δ m0 ), which don’t
appear in the naive Green function, are to correct for the fact that the figure above
is not quite the correct one: The boundaries of the initial and final strings should
not go to ±∞ before the source terms (9.2.3) (i.e., the wave functions) do, because of
the boundary conditions. The net result is that nonzero-modes appear with an extra
factor of 2 due to reflections from the boundary. However, these relative factors of 2
 π dσ
are canceled in the σ-integration, since 0 2π cos2 mσ = 14 (1 + δ m0 ). Explicitly, after
transforming the part of the ρ plane corresponding to the string to the whole of the
upper-half z plane, the Green function becomes

G(z, z  ) = ln|z − z  | + ln|z − z̄  | . (9.2.5)
9.2. Trees 181

The first term is the usual Green function without boundaries, whose use in the z
plane (and not just the ρ plane) follows from the fact that the Laplace equation is
conformally invariant. The second term, which satisfies the homogeneous Laplace
equation in the upper-half plane, has been added according to the method of images
in order to satisfy the boundary conditions at the real axis, and gives the reflections
which contribute the factors of 2.

In the latter (Schrödinger equation/operator) approach, it all boils down to using
the general expression
. .
dz  1 
dz  1
Ψ̌(z) = Ψ̌(z ) = − Ψ̌(z  ) , (9.2.6)
z 2πi z  − z r zr 2πi z  − z

where z r are the points in the z-plane representing the ends of the strings (at ρ =
±∞). Ψ̌(z) is an arbitrary operator which has been conformally transformed to the
z plane:
 w   ∞
∂ρ ρ̃r
Ψ̌r (z) = Ψ̃r (ρ) , Ψ̃r (ρ) = (p+r )−w Ψr , Ψr (ζ) = ψ rn e−nζ ,
∂z p+r −∞

ρ̃r = ρ − iπ p+s , (9.2.7b)

with Ψ(iσ) = Ψ̂(σ) in terms of P (σ) (so the ψ n ’s are the αn ’s of (7.1.7a)), and
the conformal transformations (9.2.7a) (cf. (8.1.6)) are determined by the conformal
weights w (= 1 for P̌ ). The ρ → z map is the map from the above figure to the
upper-half plane. The ζ → ρ map is the map from the free-string σ ∈ [0, π] to the rth
!r−1 !r
interacting-string σ ∈ [π s=1 p+s , π s=1 p+s ]. All σ integrals from −π to π become
contour integrals in the z plane. (The upper-half z plane corresponds to σ ∈ [0, π],
while the lower half is σ ∈ [−π, 0].) Since the string (including its extension to [−π, 0])
is mapped to the entire z plane without boundaries, (9.2.6) gets contributions from
only the ends, represented by z r . We work directly with P (σ), rather than X(σ),
since P depends only on ρ, while X depends on both ρ and ρ̄. (X̂ has a cut in the
z plane, since pσ isn’t periodic in σ.) The open string results can also be applied
directly to the closed string, which has separate operators which depend on z̄ instead
of z (i.e., + and − modes, in the notation of sect. 6.2). Actually, there is a bit of a
cheat, since P (σ) doesn’t contain the zero-mode x. However, this zero-mode needs
special care in any method: Extra factors of 2 appeared in the path-integral approach
because of the boundary conditions along the real z-axis and at infinity. In fact, we’ll
see that the lost zero-mode terms can be found from the same calculation generally
182 9. GRAPHS

used in both operator and path integral approaches to find the integration measure
of step 3 above, and thus requires no extra effort.

We therefore look directly for a propagator e∆ that gives

Ψ̌(z) = e∆ Ψ̌r (z)e−∆ = Ψ̌r (z) + [∆, Ψ̌r (z)] + · · · , (9.2.8)

where Ψ̌r corresponds to a free in-field for the rth string in the interaction picture,
˜ for which
and Ψ̌ to the interacting field. To do this, we first find a ∆

˜ Ψ̌r ] = Ψ̌ .
[∆, (9.2.9)

˜ (external-line amputation):
We next subtract out the free part of ∆

˜ = ∆ + ∆f ree
∆ , [∆f ree , Ψ̌r ] = Ψ̌r . (9.2.10)

This gives a ∆ which is quadratic in operators, but contains no annihilation operators
(which are irrelevant anyway, since the |0
will kill them). As a result, there are no
terms with multiple commutators in the expansion of the exponential. We therefore
obtain (9.2.8). When we subtract out free parts below, we will include the parts of
the external-line amputation which compensate for the fact that z and z r are not at
the same time.

We first consider applying this method to operators of arbitrary conformal weight,
as in (8.1.23). The desired form of ∆ which gives (9.2.8) for both f and δ/δf is
. dz
dz  1 ˇ δ
∆0 = − f r (z) − f ree-string terms , (9.2.11a)
r,s zr 2πi zs 2πi z − z  δ fˇs (z  )

δ δ
ˇ , fˇs (z 2 ) = 2πiδ rs δ(z 2 − z 1 ) ↔ , fˆs (σ 2 ) = 2πδ rs δ(σ 2 − σ 1 ) ,
δ f r (z 1 ) δ fˆr (σ 1 )
as follows from the fact that the conformal transformations preserve the commutation
relations of f and δ/δf . The integration contours are oriented so that
.  πp+r
dρ dσ
= =1 . (9.2.12)
ρr 2πip+r −πp+r 2πp+r

(9.2.11) can easily be shown to satisfy (9.2.8). The value of τ r (→ ±∞) for the
integration contour is fixed, so the δ in ζ in these commutation relations is really just
a δ in σ of that contour. The free-string terms are subtracted as explained above.
9.2. Trees 183

(In fact, they are poorly defined, since the integration contours for r = s fall on top
of each other.)
Unfortunately, ∆ will prove difficult to evaluate in this form. We therefore rewrite
it by expressing the 1/(z − z  ) as the derivative with respect to either z or z  of a ln
and perform an integration by parts. The net result can be written as
. dz
∆(Ψ̌1 , Ψ̌2 ) = ln(z − z  )Ψ̌1r (z)Ψ̌2s (z  )
r,s zr 2πi zs 2πi
− f ree-string terms + (zero-mode)2 terms , (9.2.13a)

[Ψ̌2r (z 2 ) , Ψ̌1s (z 1 )} = −2πiδ rs δ  (z 2 − z 1 ) , (9.2.13b)

The on the contour integration is because the integration is poorly defined due
to the cut for the ln: We therefore define it by integration by parts with respect to
either z or z  , dropping surface terms. This also kills the constant part of the ln which
contributes the (zero-mode)2 terms, which we therefore add back in. Actually, (9.2.11)
has no (zero-mode)2 terms, but in the case Ψ1 = Ψ2 = P , these terms determine
the evolution of the zero-mode x, which doesn’t appear in P , and thus could not be
determined by (9.2.8) anyway. (x does appear in X and X̂, but they’re less convenient
to work with, as explained above.) These (quadratic-in-)zero-mode contributions are
most easily calculated separately by considering the case when all external states are
ground states of nonvanishing momentum (see below). In order for the commutation
relations (9.2.13b) to be preserved by the conformal transformations, it’s necessary
that the conformal weights w 1 and w 2 of Ψ1 and Ψ2 both be 1. In that case, Ψ̌ can
be replaced with Ψ̃ in (9.2.13a) while replacing dz with dρ. However, one important
use of this equation is for the evaluation of vertices (S-matrices with no internal
propagators). Since these vertices are just δ functionals in the string field coordinates
(see below), and δ functionals are independent of conformal weight except for the
ζ → ρ transformation (since that transformation appears explicitly in the argument
of the δ functionals), we can write this result, for the cases of S-matrices with w 1 =
w 2 = 1 or vertices (with w 1 + w 2 = 2) as
. dρ . dρ
∆(Ψ̌1 , Ψ̌2 ) = ln(z − z  )Ψ̌1r (ρ)Ψ̌2s (ρ )
r,s ρr 2πi ρs 2πi
− f ree-string terms + (zero-mode)2 terms , (9.2.14a)

[Ψ2r (ζ 2) , Ψ1s (ζ 1 )} = −2πiδ rs δ  (ζ 2 − ζ 1 ) , (9.2.14b)
where the appropriate ∆ for X (w 1 = w 2 = 1) is

∆ = 12 ∆(P̌ , P̌ ) . (9.2.15)
184 9. GRAPHS

These and free-string corrections may seem awkward, but they will automatically
be fixed by the same method which gives a simple evaluation of the contour integrals:
i.e., the terms which are difficult to evaluate are exactly those which we don’t want.
(For non-vertex S-matrices with w 1 = 1 = w 2 , (9.2.11) or (9.2.13) can be used, but
their evaluated forms are much more complicated in the general case.) In general
(for covariant quantization, supersymmetry, etc.) we also need extra factors which
are evaluated at infinitesimal separation from the interaction (splitting) points, which
follow from applying the conformal transformation (9.2.7), and (9.2.6) with z = z IN T .
Only creation operators contribute.
The contour-integral form (9.2.14) can also be derived from the path-integral form
(9.2.4). For the open string [9.5], these contour integrals can be obtained by either
combining integrals over semicircles in the upper-half z plane (σ integrals from 0 to
π) with their reflections [9.6], or more directly by reformulating the open string as
a closed string with modes of one handedness only, and with interactions associated
with just the points σ = 0, π rather than all σ.
For the second step, Grs mn unfortunately is hard to calculate in general. For
open-string tree graphs, we perform the following conformal mapping to the upper-
half complex plane [9.4], where ∆ is easy to calculate:

ρ= p+r ln(z − z r ) . (9.2.16)

The boundary of the (interacting) string is now the real z axis, and the interior is
the upper half of the complex z-plane. (The branches in the ln’s in (9.2.16) are thus
chosen to run down into the lower-half plane. When we use the whole plane for
contour integrals below, we’ll avoid integrals with contours with cuts inside them.)
As a result, operators such as P , which were periodic in σ, are now meromorphic
at z r , so the contour integrals are easy to evaluate. Also, extending σ from [0, π]
to [−π, π] extends the upper-half plane to the whole complex plane, so there are no
boundary conditions to worry about. To evaluate (9.2.14), we note that, since we are
neglecting (zero-modes)2 and free string terms (r = s, m = −n), we can replace
 −mρ/p+r −nρ /p+s 1 ∂ ∂ 
ln(z − z )e → m n +  ln(z − z  ) e−mρ/p+r −nρ /p+s
+ p+s
∂ρ ∂ρ
by integration by parts. We then use the identity, for the case of (9.2.16),
∂ ∂ N
∂ ∂
+  ln(z − z  ) = p+r ln(z − z r ) ln(z  − z r ) . (9.2.18)
∂ρ ∂ρ r=2 ∂ρ ∂ρ
9.2. Trees 185

Then we can trivially change completely to z coordinates by using dρ∂/∂ρ = dz∂/∂z,
and converting the ρ exponentials into products of powers of z monomials. Differen-
tiating the ln’s gives (products of) single-variable contour integrals which can easily
be evaluated as multiple derivatives:
1 N
∆= ψ 1rm ψ 2sn (p+r )1−w1 (p+s )1−w2 p+t Artm Astn + (zero-mode)2 ,
rsmn np+r + mp+s t=2
 −m
. "
r−1 "N
dz 1 
Artm = (z − z r ) (z s − z)p+s /p+r (z − z s )p+s /p+r  .
zr 2πi z − z t s=2 s=r+1
For the third step, for open-string trees, we also need the Jacobian from dτ i →
( dz i )V (z), which for trees can easily be calculated by considering the graph where
all external states are tachyons and all but 2 strings (one incoming and one outgoing)
have infinitesimal length. We can also restrict all transverse momenta to vanish, and
determine dependence on them at the end of the calculation by the requirement of
Lorentz covariance. (Alternatively, we could complicate the calculation by including
transverse momenta, and get a calculation more similar to that of (9.1.9).) We then
have the amplitude (from nonrelativistic-style quantum mechanical arguments, or
specializing (9.2.2))

A=g N−2
f (p+r ) dτ i e− r=2
p−r τ r
, (9.2.20)

where f is a function to be determined by Lorentz covariance, the τ ’s are the inter-
action points, the strings 1 and N are those whose length is not infinitesimal, and
we also choose z 1 = ∞, z N = 0 in the transformation (9.2.16). We then solve for τ r
(= −Re(ρr )), in terms of z r (in this approximation of infinitesimal lengths for all but
2 of the strings), as the finite values of ρ where the boundary “turns around”:

 =0 →
∂z ρr
     2 
ρr = p+N ln z r + p+r ln − zr − 1 + p+s ln(z r − z s ) + O   .
p+N s=2,s=r
We then find, using the mass-shell condition p− = 1/p+ for the tachyon (p− is −H in
nonrelativistic-style calculations)
 N−1   N−1 
" p+r " "
A=g N−2
f p+N dz i z 2 (z r − z s )p+r /p+s +p+s/p+r ·
r=2 e i=3 s>r=2
186 9. GRAPHS

pir nrs (p+ ,z)pis
·e , (9.2.22a)

where we have now included the transverse-momentum factor nrs , whose exponential
form follows from previous arguments. Its explicit value, as well as that of f , can
now be determined by the manifest covariance of the tachyonic amplitude. However,
(9.2.22a) is also the correct measure for the z-integration to be applied to (9.2.1) (or
(9.2.4)), using ∆ from (9.2.19) (which is expressed in terms of the same transformation
(9.2.16)). At this point we can see that Lorentz covariance determines f to be such
that the factor in brackets is a constant. We then note that, using p− = −H =
−( 12 pi 2 + N − 1)/p+ , we have pr · ps = pir pis − [(p+r /p+s )( 12 pis 2 + N − 1) + r ↔ s].
This determines the choice of nrs which makes the amplitude manifestly covariant for

N 1 1
2 −1)−(p 2 −1)
dz V (z) = g N−2
dz i z 2 (z r − z s )pr ·ps −(p+r /p+s )( 2 ps +s /p+r )( 2 pr
i=3 s>r=2
Note that p− dependence cancels, so pr · ps and pr can be chosen to be the covariant
ones. Taking N=1 to compare with the tachyonic particle theory, we see this agrees
with the result (9.1.15) (after choosing also z 2 = 1). It also gives the (zero-mode)2
terms which were omitted in our evaluation of ∆. (That is, we have determined both
of these factors by considering this special case.) For the case of the tachyon, we
could have obtained the covariant result (9.2.22b) more directly by using covariant
amputation factors, i.e., by using p− as an independent momentum instead of as the
hamiltonian (see sect. 2.5). However, the result loses its manifest covariance, even
on shell, for excited states because of the usual 1/p+ interactions which result in the
light-cone formalism after eliminating auxiliary fields.

As mentioned in sect. 8.1, there is an Sp(2) invariance of free string theory. In
terms of the tree graphs, which were calculated by performing a conformal map to the
upper-half complex plane, it corresponds to the fact that this is the subgroup of the
conformal group which takes the upper-half complex plane to itself. Explicitly, the
transformation is z → (m11 z + m12 )/(m21 z + m22 ), where the matrix mij is real, and
without loss of generality can be chosen to have determinant 1. This transformation
also takes the real line to itself, and when combined with (9.2.16) modifies it only
by adding a constant and changing the values of the z r (but not their order). In
particular, the 3 arbitrary parameters allow arbitrary values (subject to ordering) for
z 1 , z 2 , and z N , which were ∞, 1, and 0. (This adds a term for z 1 to (9.2.16) which was
previously dropped as an infinite constant. ρ → ∞ as z → ∞ in (9.2.16) corresponds
9.2. Trees 187

to the end of the first string.) Because of the Sp(2) invariance, (9.2.22) can be
rewritten in a form with all z’s treated symmetrically: The tachyonic amplitude is
dz r "
A=g N−2 r=1
(z i − z j )(z i − z k )(z j − z k ) (z r − z s )pr ·ps , (9.2.22c)
dz i dz j dz k 1≤r<s≤N

where z i , z j , z k are any 3 z’s, which are not integrated over, and with all z’s cyclically
ordered as in (9.1.14).

Closed-string trees are similar, but whereas open-string interaction points occur
anywhere on the boundary, closed-string interaction points occur anywhere on the
surface. (Light-cone coordinates are chosen so that these points always occur for
those values of τ where the strings split or join.) Thus, for closed strings there are
also integrations over the σ’s of the interaction points. The amplitude corresponding
to (9.1.15) or (9.2.22) for the closed string, since it has both clockwise and counter-
clockwise modes, has the product of the integrand for the open string (for one set
of modes) with its complex conjugate (for the modes propagating in the opposite
direction), and the integral is over both z’s and z̄’s (i.e., both τ ’s and σ’s). (There
are also additional factors of 1/4 in the exponents due to the different normalization
of the zero modes.) However, whereas the integral in (9.1.15) for the open string is
restricted by (9.1.14) so that the z’s (interaction points) lie on the boundary (the real
axis), and are ordered, in the closed string case the z’s are anyplace on the surface
(arbitrary complex).

We next consider the evaluation of the open-string 3-point function, which will be
needed below as the vertex in the field theory action. The 3-string vertex for the open
string can be written in functional form as a δ-functional equating the coordinates
of a string to those of the strings into which it splits. In the case of general string
coordinates Z:

S IN T = g dτ D3 Z d3 p+ δ p+ δ[Z̃1 (σ) − Z̃3 (σ)]δ[Z̃2 (σ) − Z̃3 (σ)]Φ[1]Φ[2]Φ[3] ,
with Z̃ as in (9.2.7). We now use (9.2.6-16) for N= 3, with z r = ∞, 1, 0 in (9.2.16),
and p+1 with the opposite sign to p+2 and p+3 . The splitting point is

∂ρ p+3
=0 → z = z0 = − , ρ = τ 0 + iπp+2 , (9.2.24a)
∂z p+1

τ0 = 1
2 p+ ln (p+ 2 ) . (9.2.24b)
188 9. GRAPHS

For (9.2.19), we use the integral
dz −n−1 1 dn u(u − 1) · · · (u − n + 1) u
z (z + 1)u = n
(z + 1)u |z=0 = = ,
0 2πi n! dz n! n
−u + n − 1 n
= (−1) (9.2.25)
n n
to evaluate
1 −m p+,r+1
m>0: Ar2m = p+3 N rm , Ar3m = −p+2 N rm ; N rm = p+r
p+,r+1 m
m=0: Art0 = δ rt − δ r1 . (9.2.26)

The result is then (see, e.g., [9.4]):

1 1  p2 + M 2
∆(Ψ̌1 , Ψ̌2 ) = −ψ 1 Nψ 2 − ψ1N ψ2 − ψ1 N ψ2 − τ 0 ,
n n 2p+
p+1 p+2 p+3
Nrsmn = N rm N sn , ψ = p+[r (ψ 0 )r+1] ,
np+r + mp+s

S= d 3 p+ d 3 ψ δ p+ δ ψ0 Φ1 Φ2 Φ3 e∆  0 . (9.2.27)

(In some places we have used matrix notation with indices r, s = 1, 2, 3 and m, n =
1, 2, ..., ∞ implicit.) The ψ’s include the p’s. For simplicity, we have assumed the
ψ’s have w = 1; otherwise, each ψ should be replaced with p+ 1−w ψ. The τ 0 term
comes from shifting the value of τ at which the vertex is evaluated from τ = 0 to the
interaction time τ = τ 0 (it gives just the propagator factor e−τ 0 Hr
, where H r is the
free hamiltonian on each string). In more general cases we’ll also need to evaluate
 at the splitting point, which is also expressed in terms of the mode
a regularized Ψ
expansion of ln(z − z r ) (actually its derivative 1/(z − z r )) which was used in (9.2.14)
to obtain (9.2.27):
1 √
Ψ̌(z 0 ) → √ ψ + p+1 p+2 p+3 p+ −w N ψ . (9.2.28)
p+1 p+2 p+3

(Again, matrix notation is used in the second term.) We have arbitrarily chosen a
convenient normalization factor in the regularization. (A factor which diverges as
z → z 0 must be divided out anyway.) The vertex is cyclically symmetric in the
3 strings (even though some strings have p+ < 0). Besides the conservation law
! !
p = 0, we also have p+ x = 0, which is actually the conservation law for angular
momentum J +i . These are special cases of the ψ 0 conservation law indicated above by
the δ function, after including the p+ 1−w . (Remember that a coordinate of weight w is
9.2. Trees 189

conjugate to one with weight 1 − w.) This conservation law makes the definition of ψ̃
above r-independent. However, such conservation laws may be violated by additional
vertex factors (9.2.28).

The 3-string vertex for the closed string in operator form is essentially just the
product of open-string vertices for the clockwise and counterclockwise modes, since
the δ functionals can be written as such a product, except for the zero modes. How-
ever, whereas open strings must join at their ends, closed strings may join anywhere,
and the σ parametrizing this joining is then integrated over. Equivalently, the vertex
may include projection operators δ ∆N,0 = 2π
eiσ∆N which perform a σ translation
equivalent to the integration. (The former interpretation is more convenient for a
first-quantized approach, whereas the latter is more convenient in the operator for-
malism.) These projection operators are redundant in a “Landau gauge,” where the
residual σ → σ + constant gauge invariance is fixed by introducing such projectors
into the propagator.

In the covariant first-quantized formalism one can consider more general gauges
for the σ-τ reparametrization invariance and local scale invariance than g mn = η mn .
Changing the gauge has the effect of “stretching” the surface in σ-τ space. Since the
2D metric can always be chosen to be flat in any small region of the surface, it’s clear
that the only invariant quantities are global. These are topological quantities (some
integers describing the type of surface) and certain proper-length parameters (such
as the proper-length of the propagator in the case of the particle, as in (5.1.13)). In
particular, this applies to the light-cone formalism, which is just a covariant gauge
with stronger gauge conditions (and some variables removed by their equations of
motion). Thus, the planar light-cone tree graph above is essentially a flat disc, and
the proper-length parameters are the τ i , i = 3, . . . , N − 1. However, there are more
general covariant gauges even for such surfaces with just straight-line boundaries: For
example, we can identify
190 9. GRAPHS

1 3
2 4

with the proper-length parameter being the relative position of the 2 splitting points
(horizontal or vertical displacement, respectively for the 2 graphs, with the value of
the parameter being positive or negative). More generally, the only invariants in
this graph are the proper length distances measured along the boundary between
the endpoints (the points associated with the external particles), less 3 which can be
eliminated by remaining projective invariance (consider, e.g., the surface as a disc,
with the endpoints on the circular boundary). Thus, we can keep the splitting points
in the positions in the figures and vary the proper-length parameters by moving the
ends instead. If this is interpreted in terms of ordinary Feynman graphs, the first
graph seems to have intermediate states formed by the collision of particles 1 and 2,
while the second one is from 1 and 3. The identity of these 2 graphs means that the
same result can be obtained by summing over intermediate states in only 1 of these
2 channels as in the other, as we saw for the case of external tachyons in (9.1.16).
Thus, duality is just a manifestation of σ-τ reparametrization invariance and local
scale invariance.

9.3. Loops

Here we will only outline the procedure and results of loop calculations (for details
see [0.1,1.3-5,9.7-10,5.4] and the shelf of this week’s preprints in your library). In the
first-quantized approach to loops the only essential difference from trees is that the
topology is different. This means that: (1) It’s no longer possible to conformally map
to the upper-half plane, although one can map to the upper-half plane with certain
lines identified (e.g., for the planar 1-loop graph, which is topologically a cylinder, we
can choose the region between 2 concentric semicircles, with the semicircles identified).
(2) The integration variables include not only the τ ’s of the interaction points which
9.3. Loops 191

define the position of the loop in the string, but also the σ’s, which are just the p+ ’s
of the loop. In covariant gauges it’s also necessary to take the ghost coordinates
into account. In the second-quantized approach the loop graphs follow directly from
the field theory action, as in ordinary field theory. However, for explicit calculation,
the second-quantized expressions need to be translated into first-quantized form, as
for the trees. 1-loop graphs can also be calculated in the external field approach by
“sewing” together the 2 ends of the string propagator, converting the matrix element
in (9.1.9) into a trace, using the trace operator in (9.1.12).

An interesting feature of open string theories is that closed strings are generated
as bound states. This comes from stretching the one-loop graph with intermediate
states of 2 180◦ -twisted open strings:




' $

' $


& %

& %


Thus, a closed string is a bound state of 2 open strings. The closed-string coupling
can then be related to the open-string coupling, either by examining more general
192 9. GRAPHS

graphs, or by noticing that the Gauss-Bonnet theorem says that twice the number of
“handles” (closed-string loops) plus the number of “windows” (open string loops) is
a topological invariant (the Euler number, up to a constant), and thus 1 closed-string
loop can be converted into 2 open-string loops. Specifically,

h̄g closed = (h̄g open )2 → g closed = h̄g open 2 . (9.3.1)

Thus, for consistent h̄ counting the open strings must be thought of as fundamental is
such a theory (which so far means just the SO(32) superstring), and the closed strings
as bound states. Since (known) closed strings always contain gravitons, this makes the
SO(32) superstring the only known example of a theory where the graviton appears
as a bound state. The graviton propagator is the result of ultraviolet divergences due
to particles of arbitrarily high spin which sum to diverge only at the pole:
2 2 2 2 1 4 4 4 2 p2 k 2 1
dk (1 − α p k + 2α p k − · · ·) = dk 2 e−α = . (9.3.2)
0 α2 p2
(In general, p2 + M 2 appears instead of p2 , so the entire closed-string spectrum is
As mentioned in the introduction, the topology of a 2D surface is defined by a
few integers, corresponding to, e.g., the number of holes. By choosing the coordinates
of the surface appropriately (“stretching” it in various ways), the surface takes the
form of a string tree graph with one-loop insertions, each one loop insertion having
the value 1 of one of the topological invariants (e.g., 1 hole). (Actually, some of these
insertions are 1-loop closed-string insertions, and therefore are counted as 2-loop in
an open-string theory due to (9.3.1).) For example, a hole in an open-string sheet
may be pushed around so that it represents a loop as in a box graph, a propagator
correction, a tadpole, or an external line correction. Such duality transformations
can also be represented in Feynman graph notation as a consequence of the duality
properties of simpler graphs such as the 4-point tree graph (9.1.16):

@ = →
9.3. Loops 193

@ '$
= @ =
& %@


@ @
@ = @ &%
@ @
@ @
@ @
@ @

By doing stretching of such planar graphs out of the plane, these loops can even be
turned into closed-string tadpoles:


Stretching represents continuous world-sheet coordinate transformations. How-
ever, there are some coordinate transformations which can’t be obtained by combining
infinitesimal transformations, and thus must be considered separately in analyzing
gauge fixing and anomalies [9.11]. The simplest example is for a closed-string loop
(vacuum bubble), which is a torus topologically. The group of general coordinate
transformations has as a subgroup conformal transformations (which can be obtained
as a residual gauge invariance upon covariant gauge fixing, sect. 6.2). Conformal
transformations, in turn, have as a subgroup the (complex) projective group Sp(2,C):
The defining representation of this group is given by 2×2 complex matrices with de-
terminant 1, so the corresponding representation space consists of pairs of complex
194 9. GRAPHS

numbers. If we consider the transformation property of a complex variable which is
the ratio of the 2 numbers of the pair, we then find:
z1 a b z1 z1
= , z0 = (9.3.3a)
z2 c d z2 z2

az 0 + b
→ , z0 = (9.3.3b)
cz 0 + d
where ad − bc = 1. Finally, the projective group has as a discrete subgroup the
“modular” group Sp(2,Z), where a, b, c, d are (real) integers (still satisfying ad − bc =
1). To see how this relates to the torus, define the torus as the complex plane with
the identification of points
z → z + n1 z 1 + n2 z 2 (9.3.4)

for any integers n1 , n2 , for 2 particular complex numbers z 1 , z 2 which point in different
directions in the complex plane. We can then think of the torus as the parallelogram
with corners 0, z 1 , z 2 , z 1 +z 2 , with opposite sides identified, and the complex plane can
be divided up into an infinite number of equivalent copies, as implied by (9.3.4). The
conformal structure of the torus can be completely described by specifying the value
of z 0 = z 1 /z 2 . (E.g., z 1 and z 2 both change under a complex scale transformation,
but not their ratio. Without loss of generality, we can choose the imaginary part of z 0
to be positive by ordering z 1 and z 2 appropriately.) However, if we transform (z 1 , z 2 )
under the modular group as in (9.3.3a), then (9.3.4) becomes

z → z + ni z  i , z i = g ij zj , (9.3.5a)

where g ij is the Sp(2,Z) matrix, or equivalently

z → z + ni z i , ni = nj g j i . (9.3.5b)

In other words, an Sp(2,Z) transformation gives back the same torus, since the identi-
fication of points in the complex plane (9.3.4) and (9.3.5b) is the same (since it holds
are all pairs of integers ni ). We therefore define the torus by the complex parameter
z 0 , modulo equivalence under the Sp(2,Z) transformation (9.3.3b). It turns out that
the modular group can be generated by just the 2 transformations
z0 → − and z 0 → z 0 + 1 . (9.3.6)
The relevance of the modular group to the 1-loop closed-string diagram is that the
functional integral over all surfaces reduces (for tori) to an integral over z 0 . Gauge
9.3. Loops 195

fixing for Sp(2,Z) then means picking just one of the infinite number of equivalent
regions of the complex plane (under (9.3.3b)). However, for a closed string in less than
its critical dimension, there is an anomaly in the modular invariance, and the theory
is inconsistent. Modular invariance also restricts what types of compactification are

If the 2D general coordinate invariance is not violated by anomalies, it’s then
sufficient to consider these 1-loop objects to understand the divergence structure of
the quantum string theory. However, while the string can be stretched to separate any
two 1-loop divergences, we know from field theory that overlapping divergences can’t
be factored into 1-loop divergences. This suggests that any 1-loop divergences, since
they would lead to overlapping divergences, would violate the 2D reparametrization
invariance which would allow the 1-loop divergences to be disentangled. Hence, it
seems that a string theory must be finite in order to avoid such anomalies. Conversely,
we expect that finiteness at 1 loop implies finiteness at all loops. Some direct evidence
of this is given by the fact that all known string theories with fermions have 1-loop
anomalies in the usual gauge invariances of the massless particles if and only if they
also have 1-loop divergences. After the restrictions placed by tree-level duality (which
determines the ground-state mass and restricts the open-string gauge groups to U(N),
USp(N), and SO(N)), supersymmetry in the presence of massless spin-3/2 particles,
and 1-loop modular invariance, this last anomaly restriction allows only SO(32) as
an open-string gauge group (although it doesn’t restrict the closed-string theories)

Of the finite theories, the closed-string theories are finite graph-by-graph, whereas
the open-string theory requires cancellation between pairs of 1-loop graphs, with the
exception of the nonplanar loop discussed above. The 1-loop closed-string graphs
(corresponding to 2-loop graphs in the open-string theory) are (1) the torus (“handle”)
and (2) the Klein bottle, with external lines attached. The pairs of 1-loop open-string
graphs are (1) the annulus (planar loop, or “window”) + Möbius strip (nonorientable
loop) with external open and/or closed strings, and (2) the disk + a graph with the
topology of RP2 (a disk with opposite points identified) with external closed strings.
The Klein bottle is allowed only for nonoriented closed strings, and the Möbius strip
and RP2 are allowed only for nonoriented open strings.

It should be possible to simplify calculations and give simple proofs of finiteness
by the use of background field methods similar to those which in gravity and super-
symmetry made higher-loop calculations tractable and allowed simple derivations of
196 9. GRAPHS

no-renormalization theorems [1.2]. However, the use of arbitrary background string
fields will require the development of gauge-covariant string field theory, the present
status of which is discussed in the following chapters.


(1) Generalize (9.1.3) to the spinning string, using (7.2.2) instead of 12 P 2 . Writing
2 D̂dD̂(σ) → 12 D̂dD̂(σ)+δ(σ)V, V = W +θV , show that V is determined explicitly
by W .

(2) Derive (9.1.7b).

(3) Fill in all the steps needed to obtain (9.1.15) from (9.1.6,8). Derive all parts of
(9.1.16). Derive (9.1.19).

(4) Evaluate (9.1.20) by using the result (9.1.15) with N → N + 2 (but dropping 2
dτ integrations) and letting k 0 = k N+1 = 0.

(5) Derive (9.2.18).

(6) Generalize (9.2.16) to arbitrary z 1 , z 2 , z N and derive (9.2.22c) by the method
of (9.2.20,21). Take the infinitesimal form of the Sp(2) transformation and show
that it’s generated by L0 , L±1 with the correspondence (8.1.3), where z = eiσ .

(7) Derive (9.2.27). Evaluate Grsmn of (9.2.4) using (9.2.19,26,27).


In this chapter we extend the discussion of sect. 2.1 to the string and consider
interacting contributions to the Poincaré algebra of sect. 7.1 along the lines of the
Yang-Mills case treated in sect. 2.3.
For the string [10.1,9.5,10.2], it’s convenient to use a field Φ[X i (σ), p+ , τ ], since
p+ is the length of the string. This X for σ ∈ [0, πp+ ] is related to that in (7.1.7) for
p+ = 1 by X(σ, p+ ) = X(σ/p+ , 1). The hermiticity condition on the (open-string)
field is
Φ[X i (σ), p+ , τ ] = Φ† [X i (πp+ − σ), −p+ , τ ] . (10.1a)
The same relation holds for the closed string, but we may replace πp+ − σ with just
−σ, since the closed string has the residual gauge invariance σ → σ + constant,
which is fixed by the constraint (or gauge choice) ∆NΦ = 0. (See (7.1.12). In
loops, this gauge choice can be implemented either by projection operators or by
Faddeev-Popov ghosts.) As described in sect. 5.1, this charge-conjugation condition
corresponds to a combination of ordinary complex conjugation (τ reversal) with a
twist (matrix transposition combined with σ reversal). The twist effectively acts as a
charge-conjugation matrix in σ space, in the sense that expressions involving tr Φ† Φ
acquire such a factor if reexpressed in terms of just Φ and not Φ† (and (10.1a) looks
like a reality condition for a group for which the twist is the group metric). Here
Φ is an N×N matrix, and the odd mass levels of the string (including the massless
Yang-Mills sector) are in the adjoint representation of U(N), SO(N), or USp(N) (for
even N) [10.3], where in the latter 2 cases the field also satisfies the reality condition

ηΦ = (ηΦ)* , (10.1b)

where η is the group metric (symmetric for SO(N), antisymmetric for USp(N)). The
fact that the latter cases use the operation of τ reversal separately, or, by combining
with (10.1a), the twist separately, means that they describe nonoriented strings: The
string field is constrained to be invariant under a twist. The same is true for closed
strings (although closed strings have no group theory, so the choice of oriented vs.

nonoriented is arbitrary, and η = 1 in (10.1b)). The twist operator can be defined
similarly for superstrings, including heterotic strings. For general open strings the
twist is most simply written in terms of the hatted operators, on which it acts as
Ô(σ) → Ô(σ − π) (i.e., as eiπN ). For general closed strings, it takes Ô(±) (σ) →
Ô(±) (−σ). (For closed strings, σ → σ − π is irrelevant, since ∆N = 0.)

Light-cone superstring fields [10.2] also satisfy the reality condition (in place of
(10.1b), generalizing (5.4.32)) that the Fourier transform with respect to Θa is equal
to the complex conjugate (which is the analog of a certain condition on covariant

Θa Πa p+ /2
DΘ e ηΦ[Θa ] = ηΦ[Π̄a ] * . (10.1c)

Θa has a mode expansion like that of the ghost C or the spinning string’s Ψ̂ (of
the fermionic sector). The ground-state of the open superstring is described by the
light-cone superfield of (5.4.35), which is a function of the zero-modes of all the
above coordinates. Thus, the lowest-mass (massless) sector of the open superstring
is supersymmetric Yang-Mills.

The free action of the bosonic open string is [10.1,9.5]
  ∞  ∞  
† ∂
S0 = − DX i dp+ dτ tr Φ p+ i +H Φ ,
−∞ −∞ ∂τ
 2πα p+    πp+ 
dσ 1  δ 1 2 dσ 1  2 pi 2 + M 2
H= 2 −α + X i − 1 = ( 2 P i − 1) = .
0 2π δX i 2 α −πp+ 2π 2p+
The free field equation is therefore just the quantum mechanical Schrödinger equation.
(The p+ integral can also be written as 2 0 , due to the hermiticity condition. This
form also holds for closed strings, with H the sum of 2 open-string ones, as described
in sec. 7.1.) Similar remarks apply to superstrings (using (7.3.13)).

As in the first-quantized approach, interactions are described by splitting and
joining of strings, but now the graph gets chopped up into propagators and vertices:

3 6
→ 2 5

4 7

The 3-open-string vertex is then just a δ functional setting 1 string equal to 2 others,
represented by an infinitesimal strip in the world sheet. The interaction term in
the action is given by (9.2.23) for Z = X. The 3-closed-string vertex is a similar δ
functional for 3 closed strings, which can be represented as the product of 2 open-
string δ functionals, since the closed-string coordinates can be represented as the
sum of 2 open-string coordinates (one clockwise and one counterclockwise, but with
the same zero-modes). This vertex generally requires an integration over the σ of
the integration point (since closed strings can join anywhere, not having any ends,
corresponding to the gauge invariance σ → σ+constant), but the equivalent operation
of projection onto ∆N = 0 can be absorbed into the propagators.

General vertices can be obtained by considering similar slicings of surfaces with
general global topologies [10.2,9.7]. There are 2 of order g, corresponding locally to a
splitting or joining:



The former is the 3-open-string vertex. The existence of the latter is implied by
the former via the nonplanar loop graph (see sect. 9.3). The rest are order g 2 , and
correspond locally to 2 strings touching their middles and switching halves:

@ @
@ ↔
@@ @





' $'$

& %
↔ &

& %&%

' $

& %
↔ '$

& %

The type-I (SO(32) open-closed) theory has all these vertices, but the type-IIAB and
heterotic theories have only the last one, since they have only closed strings, and
they are oriented (clockwise modes are distinguishable from counterclockwise). If the
type-I theory is treated as a theory of fundamental open strings (with closed strings
as bound states, so h̄ can be defined), then we have only the first of the order-g
vertices and the first of the order-g 2.

The light-cone quantization of the spinning string follows directly from the cor-
responding bosonic formalism by the 1D supersymmetrization described in sect. 7.2.
In particular, in such a formalism the vertices require no factors besides the δ func-
tionals [10.4]. (In converting to a non-superfield formalism, integration of the vertex
over θ produces a vertex factor.) However, the projection (7.2.5) must be put in by
hand. Also, the fact that boundary conditions can be either of 2 types (for bosons
vs. fermions) must be kept in mind.

The interactions of the light-cone superstring [10.2] are done as for the light-cone
bosonic string, but there are extra factors. For example, for the 3-open-string vertex

(the interacting contribution to p− ), we have the usual δ-functionals times
δ δ δ δ δ δ
V (p− ) = P L + 12 P̄ ab + 1
24 P̄ L C
, (10.3a)
δΘa δΘb δΘa δΘb δΘc δΘd
evaluated at the splitting point. The interacting contributions to q − are given by the
same overlap δ-functionals times the vertex factors
δ δ δ δ
V (q −a ) = , V (q̄ − a ) = 16 C abcd . (10.3b)
δΘa δΘb δΘc δΘd
(The euphoric notation for q is as in sect. 5.4 for d.) These are evaluated as in (9.2.28),
where P and δ/δΘ have weight w = 1. Their form is determined by requiring that the
supersymmetry algebra be maintained. The δ functional part is given as in (9.2.27),
but now the ∆ of (9.2.14) instead of just (9.2.15) is
1  δ
∆= 2 ∆(P̌ , P̌ ) − ∆ Θ̌ , . (10.4)
δ Θ̌

As for the bosonic string, the closed-superstring vertex is the product of 2 open-
string ones (including 2 factors of the form (10.3a) for type I or II but just 1 for
the heterotic, and integrated over σ). For the general interactions above, all order-g
interactions have a single open-string vertex factor, while all order-g 2 have 2, since
the interactions of each order are locally all the same. The vertex factor is either 1
or (10.3a), depending on whether the corresponding set of modes is bosonic or super-
symmetric. When these superstring theories are truncated to their ground states, the
factor (10.3a) keeps only the zero-mode contributions, which is the usual light-cone,
3-point vertex for supersymmetric Yang-Mills, and the product of 2 such factors (for
closed strings) is the usual vertex for supergravity.

The second-quantized interacting Poincaré algebra for the light-cone string can
be obtained perturbatively. For example,

[p− , J −i ] = 0 →

[p(2) − , J (2) −i ] = 0 , (10.5a)

[p(3) − , J (2) −i ] + [p(2) − , J (3) −i ] = 0 , (10.5b)

etc., where (n) indicates the order in fields (see sect. 2.4). The solution to (10.5a)
is known from the free theory. The solution to (10.5b) can be obtained from known
results for the first-quantized theory [1.4,10.5]: The first term represents the invari-
ance of the 3-point interaction of the hamiltonian under free Lorentz transformations.

The fact that this invariance holds only on-shell is an expected consequence of the
fact that the second term in (10.5b) is simply the commutator of the free hamiltonian
with the interaction correction to the Lorentz generator. Thus, the algebra of the
complete interacting generators closes off shell as well as on, and the explicit form of
J (3) −i follows from the expression for nonclosure given in [1.4,10.5]:

J (3) − i (1, 2, 3) = −2igX r i (σint )δ p+ ∆[X i ] , (10.6)

where “r” denotes any of the three strings and ∆ represents the usual overlap-integral
δ-functionals with splitting point σ IN T . This is the analog of the generalization of
(2.3.5) to the interacting scalar particle, where p− φ → −(1/2p+ )(pi 2 φ + φ2 ). Since
p− also contains a 4-point interaction, there is a similar contribution to J (2) −i (i.e.,
(10.6) with ∆ replaced by the corresponding 3-string product derived from the 4-
string light-cone vertex in p− , and g replaced with g 2 , but otherwise the same nor-
malization). Explicit second-quantized operator calculations show that this closes the
algebra [10.6]. Similar constructions apply to superstrings [10.7].
Covariant string rules can be obtained from the light-cone formalism in the same
way as in sect. 2.6, and p+ now also represents the string length [2.7]. Thus, from
(10.2) we get the free action, in terms of a field Φ[X a (σ), X α (σ), p+ , τ ],
  ∞  ∞  
† ∂
S0 = − a
dp+ dτ tr Φ p+ i +H Φ ,
−∞ −∞ ∂τ
dσ  δ2 δ δ 1

H= 1
2 −α a2
+ +  X a2 + X α X α  − 1 .
0 2π δX δX α δX α α
The vertex is again a δ functional, in all variables.
The light-cone formalism for heterotic string field theory has also been developed,
and can be extended to further types of compactifications [10.8].
Unfortunately, the interacting light-cone formalism is not completely understood,
even for the bosonic string. There are certain kinds of contact terms which must
be added to the superstring action and supersymmetry generators to insure lower-
boundedness of the energy (supersymmetry implies positivity of the energy) and
cancel divergences in scattering amplitudes due to coincidence of vertex operator
factors [10.9], and some of these terms have been found. (Similar problems have
appeared in the covariant spinning string formulation of the superstring: see sect.
12.2.) This problem is particularly evident for closed strings, which were thought to
have only cubic interaction terms, which are insufficient to bound the potential in
Exercises 203

a formalism with only physical polarizations. Furthermore, the closed-string bound
states which have been found to follow at one loop from open-string theories by
explicitly applying unitarity to tree graphs do not seem to follow from the light-
cone field theory rules [10.1]. Since unitarity requires that 1-loop corrections are
uniquely determined by tree graphs, the implication is that the present light-cone
field theory action is incomplete, or that the rules following from it have not been
correctly applied. It is interesting to note that the type of graph needed to give
the correct closed-string poles resembles the so-called Z-graph of ordinary light-cone
field theory [10.10], which contains a line backward-moving in x+ when the light-cone
formalism is obtained as an ultrarelativistic limit, becoming an instantaneous line
when the limit is reached.


I can’t think of any.


11.1. Closed strings

Since the gauge-invariant actions for free open strings follow directly from the
methods of sects. 3.4-5 using the algebras of chapt. 8 (for the bosonic and fermionic
cases, using either OSp(1,1|2) or IGL(1) algebras), we will consider here just closed
strings, after a few general remarks.

Other string actions have been proposed which lack the complete set of Stueck-
elberg fields [11.1], and as a result they are expected to suffer from problems similar
to those of covariant “unitary” gauges in spontaneously broken gauge theories: no
simple Klein-Gordon-type propagator, nonmanifest renormalizability, and singular-
ity of semiclassical solutions, including those representing spontaneous breakdown.
Further attempts with nonlocal, higher-derivative, or incomplete actions appeared
in [11.2]. Equivalent gauge-invariant actions for the free Ramond string have been
obtained by several groups [4.13-15,11.3]. The action of [11.3] is related to the rest
by a unitary transformation: It has factors involving coordinates which are evaluated
at the midpoint of the string, whereas the others involve corresponding zero-modes.

For open or closed strings, the hermiticity condition (10.1a) now requires that
the ghost coordinates also be twisted: In the IGL(1) formalism (where the ghost
coordinates are momenta)

Φ[X(σ), C(σ), C(σ)] 
= Φ† [X(π − σ), −C(π − σ), C(π − σ)] (11.1.1a)

(C gets an extra “−” because the twist is σ reversal, and C carries a σ index in the
mechanics action), and in the OSp(1,1|2) formalism we just extend (10.1a):

Φ[X a (σ), X α (σ), p+ ] = Φ† [X a (πp+ − σ), X α (πp+ − σ), −p+ ] . (11.1.1b)

As in the light-cone formalism (see sect. 10.1), for discussing free theories, we scale
σ by p+ in (11.1.1b), so the twist then takes σ → π − σ. (For the closed string the
11.1. Closed strings 205

twist is σ → −σ, so in both (11.1.1a) and (11.1.1b) the arguments of the coordinates
are just −σ on the right-hand side.)
In the rest of this section we will consider closed strings only. First we show how
to extend the OSp(1,1|2) formalism to closed strings [4.10]. By analogy to (4.1.1),
the kinetic operator for the closed string is a δ function in IOSp(1,1|2):

S= dD xd2 xα dx− d2 ∆pα d∆p+ Φ† p+ 2 δ(J AB )δ(∆pA ) Φ ,

δΦ = 12 J AB ΛBA + ∆pA ΛA , (11.1.2)

where, as in (7.1.17), the Poincaré generators J AB and pA are given as sums, and
∆pA as differences, of the left-handed and right-handed versions of the open-string
generators of (7.1.14). The Hilbert-space metric necessary for hermiticity is now
p+ 2 (or equivalently p+ (+) p+ (−) ), since a factor of p+ is needed for the open-string
modes of each handedness. For simplicity, we do not take the physical momenta pa
to be doubled here, since the IOSp(1,1|2) algebra closes regardless, but they can be
doubled if the corresponding δ functions and integrations are included in (11.1.2).
More explicitly, the δ function in the Poincaré group is given by

δ(J AB )δ(∆pA ) = δ(J αβ 2 )iδ(J −+ )δ 2 (J +α )δ 2 (J −α )δ(∆p− )δ 2 (∆pα )δ(∆p+ ) . (11.1.3)

To establish the invariance of (11.1.2), the fact that (4.1.1) is invariant indicates that
it’s sufficient to show that δ(J AB ) commutes with δ(∆pA ). This follows from the fact
that each of the J AB ’s commutes with δ(∆pA ). We interpret δ(∆p− ) = p+ δ(∆N)
in the presence of the other δ(∆p)’s, where δ(∆N) is a Kronecker δ, and the other
δ(∆p)’s are Dirac δ’s.
All the nontrivial terms are contained in the δ 2 (J −α ). As in the open-string case,
dependence on the gauge coordinates xα and x− is eliminated by the δ 2 (J +α ) and
δ(J −+ ) on the left, and further terms are killed by δ(J αβ 2 ). Similarly, dependence
on ∆pα and ∆p+ is eliminated by the corresponding δ functions on the right, and
further terms are killed by δ(∆p− ). For convenience, the latter elimination should be
done before the former. After making the redefinition Φ → p+
Φ and integrating out
the unphysical zero-modes, the action is similar to the OSp(1,1|2) case:

S= dD x 12 φ† δ(∆N)δ(M αβ 2 ) 2 − M 2 + (M α a pa + M αm M)2 φ ,

δφ = (M αa pa + M α m M)Λα + 12 M αβ Λαβ + ∆NΛ . (11.1.4)

This is the minimal form of the closed-string action.

The nonminimal form is obtained by analogy to the IGL(1) formalism, in the
same way OSp(1,1|2) was extended to IOSp(1,1|2): Using a δ function in the closed-
string group GL(1|1) of sect. 8.2 (found from sums and differences of the expressions
in (8.2.6), as in (7.1.17)), we obtain the action (with ∆p− → ∆N)

S= dD xdcd∆pc̃ Φ† iQδ(J 3 )δ(∆pc̃ )δ(∆N) Φ ,

δΦ = QΛ + J 3 Λ̀ + ∆N Λ̃ + ∆pc̃ Λ̌ , (11.1.5)
or, after integrating out ∆pc̃ , with Φ = φ + ∆pc̃ ψ,

S=  J3 )δ(∆N) φ ,
dD xdc φ† iQδ(
 + J3 Λ̀ + ∆N Λ̃ ,
δφ = QΛ (11.1.6)
where the  ’s indicate that all terms involving ∆pc̃ and its canonical conjugate have
been dropped. The field φ = ϕ + cχ is commuting.
For the gauge-fixing in the GL(1|1) formalism above (or the equivalent one from
first-quantization), we now choose [4.5]
∂ ∂ ∂
O = −2∆p c̃
c, → K = ∆p c̃
c(2 − M ) − 4M
2 +
− 2∆N c, ,
∂c ∂c ∂c
where we have used
∂ ∂
Q = −i 14 c(p2 + M 2 ) + i 12 M +
− i∆N + i 12 ∆M + ∆pc̃ + Q+ . (11.1.8)
∂c ∂∆pc̃
Expanding the string field over the ghost zero-modes,

Φ = (φ + icφ) + i∆pc̃ (ψ̂ + cφ̂) , (11.1.9)
1 †
we substitute into the lagrangian L = 2 Φ KΦ and integrate over the ghost zero
∂ ∂

L = 12 φ† (2 − M 2 )φ + 2ψ † M + ψ + 4i(φ̂† ∆Nφ + iψ̂ † ∆Nψ) . (11.1.10)
∂c ∂∆p
φ contains propagating fields, ψ contains BRST auxiliary fields, and φ̂ and ψ̂ contain
lagrange multipliers which constrain ∆N = 0 for the other fields. Although the
propagating fields are completely gauge-fixed, the BRST auxiliary fields again have
the gauge transformations

δψ = λ , M +λ = 0 , (11.1.11a)

and the lagrange multipliers have the gauge transformations

δ φ̂ = λ̂φ , δ ψ̂ = λ̂ψ ; ∆N λ̂φ = ∆N λ̂ψ = 0 . (11.1.11b)
11.2. Components 207

11.2. Components

To get a better understanding of the gauge-invariant string action in terms of more
familiar particle actions, we now expand the string action over some of the lower-mass
component fields, using the algebras of chapt. 8 in the formalism of sect. 4. All of
these results can also be derived by simply identifying the reducible representations
which appear in the light-cone, and then using the component methods of sect. 4.1.
However, here we’ll work directly with string oscillators, and not decompose the
reducible representations, for purposes of comparison.
As an example of how components appear in the IGL(1) quantization, the mass-
less level of the open string is given by (cf. (4.4.6))

Φ = (Aa aa † + Cα aα † ) + icBac̃† |0
, (11.2.1)

where aα = (ac , ac̃ ), all oscillators are for the first mode, and we have used (4.4.5).
The lagrangian and BRST transformations then agree with (3.2.8,11) for ζ = 1. In
order to obtain particle actions directly without having to eliminate BRST auxiliary
fields, from now on we work with only the OSp(1,1|2) formalism. (By the arguments
of sect. 4.2, the IGL(1) formalism gives the same actions after elimination of BRST
auxiliary fields.)
As described in sect. 4.1, auxiliary fields which come from the ghost sector are
crucial for writing local gauge-invariant actions. (These auxiliary fields have the same
dimension as the physical fields, unlike the BRST auxiliary fields, which are 1 unit
higher in dimension and have algebraic field equations.) Let’s consider the counting
of these auxiliary fields. This requires finding the number of Sp(2) singlets that can
be constructed out of the ghost oscillators at each mass level. The Sp(2) singlet
constructed from two isospinor creation operators is am α† anα † , which we denote as
(mn). A general auxiliary field is obtained by applying to the vacuum some nonzero
number of these pairs together with an arbitrary number of bosonic creation operators.
The first few independent products of pairs of fermionic operators, listed by eigenvalue
of the number operator N, are:

0 : I

1 : −

2 : (11)

3 : (12)

4 : (13) , (22)

5 : (14) , (23)

6 : (15) , (24) , (33) , (11)(22) (11.2.2)

where I is the identity and no operator exists at level 1. The number of independent
products of singlets at each level is given by the partition function
n=2 (1 − x )

= 1+x2 +x3 +2x4 +2x5 +4x6 +4x7 +7x8 +8x9 +12x10 +14x11 +21x12 +· · · , (11.2.3)

corresponding to the states generated by a single bosonic coordinate missing its zeroth
and first modes, as described in sect. 8.1.

We now expand the open string up to the third mass level (containing a massive,
symmetric, rank-2 tensor) and the closed string up to the second mass level (contain-
ing the graviton) [4.1]. The mode expansions of the relevant operators are given by
(8.2.1). Since the δ(M αβ 2 ) projector keeps only the Sp(2)-singlet terms, we find that
up to the third mass level the expansion of φ is

φ = [φ0 + Aa a† 1a

+ 12 hab a† 1a a† 1b + B a a† 2a + η(a1α † )2 ] |0
. (11.2.4)

Here φ0 is the tachyon, Aa is the massless vector, and (hab , B a , η) describe the massive,
symmetric, rank-two tensor. It’s now straightforward to use (8.2.1) to evaluate the
action (4.1.6):
L = L−1 + L0 + L1 ; (11.2.5)

L−1 = 12 φ0 (2 + 2)φ0 , (11.2.6a)

L0 = 12 A · 2 A + 12 (∂ · A)2 = − 14 F 2 , (11.2.6b)

L1 = 14 hab (2 − 2)hab + 12 B · (2 − 2)B − 12 η(2 − 2)η

+ 12 (∂ b hab + ∂ a η − B a )2 + 12 ( 14 ha a + 32 η + ∂ · B)2 . (11.2.6c)

The gauge transformations are obtained by expanding (4.1.6). The pieces involv-
ing Λαβ are trivial in the component viewpoint, since they are the ones that reduce
the components of φ to the Sp(2) singlets given in (11.2.4). Since then only the Sp(2)
11.2. Components 209

singlet part of Qα Λα can contribute, only the (M αβ ) isospinor sector of Λα is relevant.
We therefore take
Λα = (ξa1α † +
a a1a † a1α † +
a2α † ) |0
. (11.2.7)

Then the invariances are found to be

δAa = ∂ a ξ ; (11.2.8a)

δhab = ∂ (a
b) − √ η ab

δB a = ∂ a
+ 2
a ,
δη = −∂ ·
+ √
. (11.2.8b)
(11.2.6b) and (11.2.8a) are the usual action and gauge invariance for a free photon;
however, (11.2.6c) and (11.2.8b) are not in the standard form for massive, symmetric
rank two. Letting

hab = h 1  a −
η = − 12 h 3 
ab 10 η ab η , a 10 η ; (11.2.9)

one finds
√ √

δh , δB a = ∂ a
+ 2
a , δ η = −5 2
. (11.2.10)
ab (a b)

In this form it’s clear that η and B a are Stueckelberg fields that can be gauged away
a . (This was not possible for η, since the presence of the ∂ ·
term in
its transformation law (11.2.8b) would require propagating Faddeev-Popov ghosts.)
In this gauge L1 reduces to the Fierz-Pauli lagrangian for a massive, symmetric,
rank-two tensor:

 ab 2 h
b ab ) − 2 (∂ b h )∂ a h c − 4 h a 2 h b − 4 (h hab − h a ) . (11.2.11)
L = 14 h  + 1 (∂ h
 2 1  ab c 1 a b 1  ab  a 2
ab 2

The closed string is treated similarly, so we’ll consider only the tachyon and
massless levels. The expansion of the physical closed-string field to the second mass
level is

φ = (φ0 + hab a+ 1a † a− 1b † + Aab a+ 1a † a− 1b † + ηa+ 1 α† a− 1α † ) |0
, (11.2.12)

where φ0 is the tachyon and hab , Aab , and η describe the massless sector, consisting
of the graviton, an antisymmetric tensor, and the dilaton. We have also dropped

fields which are killed by the projection operator for ∆N. We then find for the action

L = L−2 + L0 ;

L−2 = 12 φ0 (2 + 4)φ0 ,

L0 = 14 hab 2 hab + 14 Aab 2 Aab − 12 η 2 η

+ 12 (∂ b hab + ∂ a η)2 + 12 (∂ b Aab )2 . (11.2.13)

The nontrivial gauge transformations are found from (11.1.4):

δhab = ∂ (a
b) , δAab = ∂ [a ζ b] , δη = ∂ ·
. (11.2.14)

These lead to the field redefinitions

 + η η ,
hab = h a
η = η + 12 h ; (11.2.15)
ab ab a

which result in the improved gauge transformations


δh , δAab = ∂ [a ζ b] , δ η = 0 . (11.2.16)
ab (a b)

Substituting back into (11.2.13), we find the covariant action for a tachyon, linearized
Einstein gravity, an antisymmetric tensor, and a dilaton [4.10].

The formulation with the world-sheet metric (sect. 8.3) uses more gauge and
auxiliary degrees of freedom than even the IGL(1) formulation. We begin with the
open string [3.13]. If we evaluate the kinetic operator for the gauge-invariant action
 2 → !∞ (b † O +
(4.1.6) between states without fermionic oscillators, we find Q 1 n n

On † bn ), so the kinetic operator reduces to

δ(M αβ 2  2)
)(−2K + Q → −2K(1 − N GB ) + †  +L
−bn bn + √ (bn † L  †b ) ,
n n n
1 n

N GB = (bn † g n + g n † bn ) , (11.2.17)
 . The operator N
dropping the f and c terms in K and L †
n GB counts the number of g ’s

plus b† ’s in a state (without factors of n). We now evaluate the first few component
levels. The evaluation of the tachyon action is trivial:

φ = ϕ(x) |0
→ L = 12 ϕ(2 + 2)ϕ , (11.2.18)
11.2. Components 211

where S = d26 x L. For the photon, expanding in only Sp(2) singlets,

φ = (A · a1 † + Bg 1 † + Gb1 † ) |0

L = 12 A · 2 A − 12 B2 − B∂ · A = − 14 F ab 2 − 12 (B + ∂ · A)2 . (11.2.19)

The disappearance of G follows from the gauge transformations

Λα = (f 1α † λc + c1α † λf ) |0

δ(A, B, G) = (∂λc , −2 λc , λf − 12 λc ) . (11.2.20)

Since G is the only field gauged by λf , no gauge-invariant can be constructed from
it, so it must drop out of the action.

For the next level, we consider the gauge transformations first in order to deter-
mine which fields will drop out of the action so that its calculation will be simplified.
The Sp(2) singlet part of the field is

φ = ( 12 hab a1a † a1b † + ha a2a † + Ba g 1 † a1a † + Ga b1 † a1a †
 †2 + Gb †2 + Bg † + Gb
+ hg 1 † b1 † + Bg  † + η f †2 + η c †2 + η f α† c † ) |0
1 1 2 2 + 1 − 1 0 1 1α


The only terms in Λα which contribute to the transformation of the Sp(2) singlets are

Λα = (λcp · a† 1 + λcb g †1 + λcg b† 1 )f † 1α + (λf p · a† 1 + λf b g † 1 + λf g b† 1 )c† 1α
† † 
+ λc f 2α + λf c 2α 0 . (11.2.22)

The gauge transformations of the components are then
δhab = ∂ (a λcp b) − η ab √ λc ,
√ cp
δha = 2λ a + ∂ a λc ,

δBa = ∂ a λcb + 2Kλcp a ,

δGa = − 12 λcp a + λf p a + ∂ a λcg ,
δh = √ λc − 12 λcb + λf b + 2Kλcg ,
 = 2Kλ
δB ,

δG = − 12 λcg + λf g ,

δB = 2λcb + 2Kλc ,

 = − 1 λ + λ + 2λ
δG ,
2 c f cg

δη + = λcb ,

δη − = 2λf − 12 λf b − ∂ · λf p + 2Kλf g ,

δη 0 = 2λc − 14 λcb + 12 λf b − 12 ∂ · λcp + Kλcg . (11.2.23)

We then gauge away

Ga = 0 → λf p a = 12 λcp a − ∂ a λcg ,

G =0 → λf g = 12 λcg ,

η+ = 0 → λcb = 0 ,
1 1 1
η− = 0 → λf = √ λf b + √ ∂ · λcp − √ (2 + K)λcg ,
2 2 2 2 2

η0 = 0 → λf b = −2 2λc + ∂ · λcp − 2Kλcg . (11.2.24)

The transformation laws of the remaining fields are

δhab = ∂ (a λb) − η ab λ ,

δha = 2(λa + ∂ a λ) ,

δh = −3λ + ∂ · λ ,

δBa = −(2 − 2)λa ,

δB = − 2(2 − 2)λ ,

δB =0 ,

 = √ 1
δG (−3λ + ∂ · λ) , (11.2.25)

where λa = λcp a and λ = λc / 2. λcg drops out of the transformation law as a result
of the gauge invariance for gauge invariance
δφ = Q → βΛ
δΛα = Q . (11.2.26)
α (αβ)

The lagrangian can then be computed from (4.1.6) and (11.2.17) to be

L = 14 hab (2 − 2)hab + 12 ha (2 − 2)ha − 12 h(2 − 2)h

− 12 Ba 2 − 12 B2 + 2B(  − h)
√ 3 1
+ Ba (−∂ b hab + ∂ a h + 2ha ) + B(−∂ · h + √ h − √ ha a ) . (11.2.27)
2 2 2
11.2. Components 213

The lagrangian of (11.2.19) is equivalent to that obtained from the IGL(1) action,
whereas the lagrangian of (11.2.27) is like the IGL(1) one but contains in addition
 and 2G
the 2 gauge-invariant auxiliary fields B − h. Both reduce to the OSp(1,1|2)
lagrangians after elimination of auxiliary fields.

The results for the closed string are similar. Here we consider just the massless
level of the nonoriented closed string (which is symmetric under interchange of + and
− modes). The Sp(2) and ∆N invariant components are

φ = (hab a† +a a† −a + B a g †(+ a† −)a + Ga b† (+ a† −)a + hg † (+ b† −) + Bg † + g †− + Gb† + b† −

+ η + f † + α f † −α + η − c† + α c† −α + η 0 f † (+ α c† −)α ) |0
, (11.2.28)

where all oscillators are from the first mode (n = 1), and hab is symmetric. The gauge
parameters are

Λα = (λcp · a† + λcb g † + λcg b† )(+ f † −)α + (λf p · a† + λf b g † + λf g b† )(+ c† −)α |0
The component transformations are then

δhab = ∂ (a λcp b) ,

δB a = ∂ a λcb − 12 2 λcp a ,

δGa = 2λf p a − λcp a + ∂ a λcg ,

δh = 2λf b − λcb − 12 2 λcg ,

δB = −2 λcb ,

δG = 4λf g − 2λcg ,

δη + = 2λcb ,

δη − = −λf b − ∂ · λf p − 12 2 λf g ,

δη 0 = λf b − 12 λcb − 12 ∂ · λcp − 14 2 λcg . (11.2.30)

We choose the gauges

Ga = 0 → λf p a = 12 λcp a − 12 ∂ a λcg ,

G =0 → λf g = 12 λcg ,

η+ = 0 → λcb = 0 ,

η0 = 0 → λf b = 12 ∂ · λcp + 14 2 λcg . (11.2.31)

The remaining fields transform as

δhab = ∂ (a λb) ,

δB a = − 12 2 λa ,

δh = ∂ · λ ,

δB = 0 ,

δη − = −∂ · λ , (11.2.32)

where λa = λcp a , and λcg drops out. Finally, the lagrangian is

L = 12 hab 2 hab − h2 h − 4B a 2 − 8B(η − + h) − 4B a (∂ b hab − ∂ a h) . (11.2.33)

Again we find the auxiliary fields B and η − + h in addition to the usual nonminimal
ones. After elimination of auxiliary fields, this lagrangian reduces to that of (11.2.13)
(for the nonoriented sector, up to normalization of the fields).


(1) Derive the gauge-invariant actions (IGL(1) and OSp(1,1|2)) for free open strings.
Do the same for the Neveu-Schwarz string. Derive the OSp(1,1|2) action for the
Ramond string.

(2) Find all the Sp(2)-singlet fields at the levels indicated in (11.2.2). Separate them
into sets corresponding to irreducible representations of the Poincaré group (in-
cluding their Stueckelberg and auxiliary fields).

(3) Derive the action for the massless level of the open string using the bosonized
ghosts of sect. 8.1.

(4) Derive the action for the next mass level of the closed string after those in
(11.2.13). Do the same for (11.2.33).

(5) Rederive the actions of sect. 11.2 for levels which include spin 2 by first decompos-
ing the corresponding light-cone representations into irreducible representations,
and then using the Hilbert-space constructions of sect. 4.1 for each irreducible
12.1. Introduction 215


12.1. Introduction

The gauge-invariant forms of the interacting actions for string field theories are far
from understood. Interacting closed string field theory does not yet exist. (Although
a proposal has been made [12.1], it is not even at the point where component actions
can be examined.) An open-string formulation exists [4.8] (see the following section),
but it does not seem to relate to the light-cone formulation (chapt. 10), and has more
complicated vertices. Furthermore, all these formulations are in the IGL(1) formalism,
so relation to particles is less direct because of the need to eliminate BRST auxiliary
fields. Most importantly, the concept of conformal invariance is not clear in these
formulations. If a formulation could be found which incorporated the world-sheet
metric as coordinates, as the free theory of sect. 8.3, it might be possible to restore
conformal transformations as a larger gauge invariance which allowed the derivation
of the other formulations as (partial) gauge choices.

In this section we will mostly discuss the status of the derivation of an interacting
gauge-covariant string theory from the light cone, with interactions similar to those of
the light-cone string field theory. The derivation follows the corresponding derivation
for Yang-Mills described in sects. 3.4 and 4.2 [3.14], but the important step (3.4.17)
of eliminating p+ dependence has not yet been performed.

As performed for Yang-Mills in sect. 3.4, the transformation (3.4.3a) with Φ →
U Φ is the first step in deriving an IGL(1) formalism for the interacting string from
the light cone. Since the transformation is nonunitary, the factor of p+ in (2.4.9) is
canceled. In (2.4.7), using integration by parts, (3.4.3a) induces the transformation
of the vertex function

V (n) → U(1) · · · U(n)V (n) , (12.1.1)
p+1 · · · p+n

where we work in momentum space with respect to p+ . The lowest-order interacting
contribution to Q then follows from applying this transformation to the OSp-extended
form of the light-cone vertex (10.6):

J (3) − c = −2ig X r c (σint ) δ p+ ∆[X a ] ∆[p+ X c ] ∆[p+ −1 X c̃ ] , (12.1.2)
p+1 p+2 p+3

effectively giving conformal weight -1 to X c and conformal weight 1 to X c̃ . A δ-
functional that matches a coordinate must also match the σ-derivative of the coordi-
nate, and with no zero-modes one must have
−1 σ  σ σ
∆ p+ X c̃
 = ∆ ∂σ p+ −1 X c̃ = ∆ p+ X −2 c̃
. (12.1.3)
p+ c̃=0 p+ p+

(Even the normalization is unambiguous, since without zero-modes ∆ can be normal-
ized to 1 between ground states.) We now recognize X c and X c̃ to be just the usual
Faddeev-Popov ghost C(σ) of τ -reparametrizations and Faddeev-Popov anti ghost

δ/δ C(σ) of σ-reparametrizations (as in (8.1.13)), of conformal weights -1 and 2, re-
spectively, which is equivalent to the relation (8.2.2,3) (as seen by using (7.1.7b)).
Finally, we can (functionally) Fourier transform the antighost so that it is replaced
with the canonically conjugate ghost. Our final vertex function is therefore

J (3) − c = −2ig C r (σint )δ 
p+ ∆[X a ] ∆[p+ C] ∆[p+ C] , (12.1.4a)
p+1 p+2 p+3
or in terms of momenta

δ δ
J(3) c
− = −2igp+r C r (σint )δ −1
p+ ∆[p+ P ] ∆ p+a −2
∆ p+ −2  .
δC δC
The extra p+ ’s disappear due to Fourier transformation of the zero-modes c:
1  −cpc̃

−cpc̃ /p+ ˜ pc̃
dc e f (p+ c) = dc e f (c) = f . (12.1.5)
p+ p+

Equivalently, the exponent of U by (3.4.3a) is c∂/∂c + M 3 , so the zero-mode part
just scales c, but c∂/∂c = 1 − (∂/∂c)c, so besides scaling ∂/∂c there is an extra
factor of p+ for each zero-mode, canceling those in (12.1.1,2). There is no effect on
the normalization with respect to nonzero-modes because of the above-mentioned
normalization in the definition of ∆ with respect to the creation and annihilation
operators. A similar analysis applies to closed strings [12.2].

Hata, Itoh, Kugo, Kunitomo, and Ogawa [12.3] proposed an interacting BRST
operator equivalent to this one, and corresponding gauge-fixed and gauge-invariant
12.2. Midpoint interaction 217

actions, but with p+ treated as an extra coordinate as in [2.7]. (A similar earlier
attempt appeared in [4.9,7.5], with p+ fixed, as a consequence of which the BRST
algebra didn’t close to all orders. Similar attempts appeared in [12.4].) However,
as explained in [2.7], such a formalism requires also an additional anticommuting
coordinate in order for the loops to work (as easily checked for the planar 1-loop
graph with external tachyons [2.7]), and can lead to problems with infrared behavior
The usual four-point vertex of Yang-Mills (and even-higher-point vertices of grav-
ity for the closed string) will be obtained only after field redefinitions of the massive
fields. This corresponds to the fact that it shows up in the zero-slope limit of the
S-matrix only after massive propagators have been included and reduced to points.
In terms of the Lagrangian, for arbitrary massive fields µ and massless fields ν, the
terms, for example,

L = 12 µ[2 + M 2 + M 2 U(ν)]µ + M 2 µV (ν) (12.1.6a)

(where U(ν) and V (ν) represent some interaction terms) become, in the limit M 2 →
L = M 2 [ 12 (1 + U)µ2 + V µ] . (12.1.6b)

The corresponding field redefinition is
µ = µ̃ − , (12.1.6c)
which modifies the Lagrangian to
L = 12 µ̃(2 + M 2 + M 2 U)µ̃ + 12 M 2 + O(M 0 ) . (12.1.6d)
The redefined massive fields µ̃ no longer contribute in the zero-slope limit, and can
be dropped from the Lagrangian before taking the limit. However, the redefinition
has introduced the new interaction term 12 M 2 1+U
into the ν-part of the Lagrangian.

12.2. Midpoint interaction

Witten has proposed an extension of the IGL(1) gauge-invariant open bosonic
string action to the interacting case [4.8]. Although there may be certain limitations
with his construction, it shares certain general properties with the light-cone (and
covariantized light-cone) formalism, and thus we expect these properties will be com-
mon to any future approaches. The construction is based on the use of a vertex which

consists mainly of δ-functionals, as in the light-cone formalism. Although the geome-
try of the infinitesimal surface corresponding to these δ-functionals differs from that
of the light-cone case, they have certain algebraic features in common. In particular,
by considering a structure for which the δ-functional (times certain vertex factors)
is identified with the product operation of a certain algebra, the associativity of the
product follows from the usual properties of δ-functionals. This is sufficient to define
an interacting, nilpotent BRST operator (or Lorentz generators with [J −i , J −j ] = 0),
which in turn gives an interacting gauge-invariant (or Lorentz-invariant) action.

The string fields are elements of an algebra: a vector space with an outer product ∗.
We write an explicit vector index on the string field Φi , where i = Z(σ) is the co-
ordinates (X, C, C for the covariant formalism and X T , x± for the light cone), and
excludes group-theory indices. Then the product can be written in terms of a rank-3
(Φ ∗ Ψ)i = f i jk Φk Ψj . (12.2.1)

In order to construct actions, and because of the relation of a field to a first-quantized
wave function, we require, in addition to the operations necessary to define an algebra,
a Hilbert-space inner product

= DZ tr Φ† Ψ = tr Φi† Ψi , (12.2.2)

where tr is the group-theory trace. Furthermore, in order to give the hermiticity
condition on the field we require an indefinite, symmetric charge-conjugation matrix
Ω on this space:

Φi = Ωij Φj † , 
(ΩΦ)[X(σ), C(σ), C(σ)] 
≡ Φ[X(π − σ), C(π − σ), −C(π − σ)] .
Ω is the “twist” of (11.1.1). To allow a reality condition or, combining with (12.2.3a),
a symmetry condition for real group representations (for SO(N) or USp(N)), we also
require that indices can be freely raised and lowered:

(ηΦ)i = Ωij δ jk (ηΦ)t k , (12.2.3b)

where η is the group metric and t is the group-index transpose. In order to perform
the usual graphical manipulations implied by duality, the twist must have the usual
effect on vertices, and thus on the inner product:

Ω(Φ ∗ Ψ) = (ΩΨ) ∗ (ΩΦ) . (12.2.4)
12.2. Midpoint interaction 219

Further properties satisfied by the product follow from the nilpotence of the BRST
operator and integrability of the field equations QΦ = 0: Defining

QΦ = Q0 Φ + Φ ∗ Φ → (QΦ)i = Q0i j Φj + f i jk Φk Φj , (12.2.5)

S= DZ tr Φ† ( 12 Q0 Φ + 13 Φ ∗ Φ) = tr 1 ij
2 (ΩQ0 ) Φj Φi + 13 (Ωf )ijk Φk Φj Φi ,
we find that hermiticity requires

(ΩQ0 )ij = (Q̄0 Ω)ji , ¯
(Ωf )ijk = (fΩΩ) kji
, (12.2.7)

integrability requires antisymmetry of Q0 and cyclicity of ∗:

(ΩQ0 )ij = −(ΩQ0 )ji , (Ωf )ijk = (Ωf )jki (12.2.8)

(where permutation of indices is in the “graded” sense, but we have omitted some
signs: e.g., (ΩQ0 )ij Φj Ψi = +(ΩQ0 )ij Ψj Φi when Φ[Z] and Ψ[Z] are anticommuting,
but Φi and Ψi include components of either statistics), and nilpotence requires, besides
Q0 2 = 0, that ∗ is BRST invariant (i.e., Q0 is distributive over ∗) and associative:

Q0 (Φ ∗ Ψ) = (Q0 Φ) ∗ Ψ + Φ ∗ (Q0 Ψ)

↔ Q0i l (f ΩΩ)ljk + Q0j l (f ΩΩ)ilk + Q0k l (f ΩΩ)ijl = 0 , (12.2.9)

Φ ∗ (Ψ ∗ Υ) = (Φ ∗ Ψ) ∗ Υ ↔ (Ωf )ijm f m kl = (Ωf )jkmf m li (12.2.10)

(where we have again ignored some signs due to grading). ∗ should also be invariant
under all transformations under conserved quantities, and thus the operators ∂/∂z ∼ 

dσ ∂/∂Z must also be distributive over ∗.
At this point we have much more structure than in an ordinary algebra, and
only one more thing needs to be introduced in order to obtain a matrix algebra: an
identity element for the outer product

Φ∗I =I ∗Φ= Φ ↔ f i jk I k = δ i j . (12.2.11)

It’s not clear why string field theory must have such an object, but both the light-cone
approach and Witten’s approach have one. In the light-cone approach the identity
element is the ground state (tachyon) at vanishing momentum (including the string
length, 2παp+ ), which is related to the fact that the vertex for an external tachyon
takes the simple form : e−ik·X(0) :. We now consider the fields as being matrices in
Z(σ)-space as well as in group space, although the vector space on which such matrices

act might not be (explicitly) defined. (Such a formalism might be a consequence of
the same duality properties that require general matrix structure, as opposed to just
adjoint representation, in the group space.) ∗ is now the matrix product. (12.2.7,10)
then express just the usual hermiticity and associativity properties of the matrix
product. The trace operation T r of these matrices is implied by the Hilbert-space
inner product (12.2.2):

= T r Φ† Ψ ↔ T r Φ = I|Φ
. (12.2.12)

(12.2.8) states the usual cyclicity of the trace. Finally, the twist metric (12.2.3) is
identified with the matrix transpose, in addition to transposition in the group space,
as implied by (12.2.4). Using this transposition in combination with the usual her-
mitian conjugation to define the matrix complex conjugate, the hermiticity condition
(12.2.3a) becomes just hermiticity in the group space: Denoting the group-space
matrix indices as Φa b ,
Φa b = Φb a * . (12.2.13)

As a result of
It = I ↔ Ωij I j = I i (12.2.14)

and the fact that Q0 and ∂/∂z are distributive as well as being “antisymmetric” (odd
under simultaneous twisting and integration by parts), we find

Q0 I = I=0 . (12.2.15)

Given one ∗ product, it’s possible to define other associative products by com-
bining it with some operators d which are distributive over it. Thus, the condition of
associativity of  and ∗ implies

A  B = A ∗ dB → d2 = 0

A  B = (dA) ∗ (dB) → d2 = d or d2 = 0 . (12.2.16)

The former allows the introduction of conserved anticommuting factors (as for the
BRST open-string vertex), while the latter allows the introduction of projection op-
erators (as expected for closed strings with respect to ∆N).
The gauge transformations and action come directly from the interacting BRST
operator: Using the analysis of (4.2.17-21),
δΦ = ΛΦ, Q , Φ = Q0 Λ + [Φ, ∗Λ] , (12.2.17a)
c c
12.2. Midpoint interaction 221

where the last bracket is the commutator with respect to the ∗ product, and

1 †
S = −iQ = 2 Φ Q0 Φ + 13 Φ† (Φ ∗ Φ) , (12.2.17b)

where for physical fields we restrict to

J 3Φ = 0 , J 3 Λ = −Λ . (12.2.17c)

Gauge invariance follows from Q2 = 2 [Q, Q]c = 0. Actually, the projection onto
J 3 = 0 is somewhat redundant, since the other fields can be removed by gauge
transformations or nondynamical field equations, at least at the classical level. (See
(3.4.18) for Yang-Mills.)

A possible candidate for a gauge-fixed action can be written in terms of Q as


S = Q, 1
2 Φ OΦ + 1
6 Φ† (QIN T Φ) (12.2.18a)
† ∂ † ∂
= − 12 (Q0 Φ) c, Φ− 1
2 (QIN T Φ) c, − 1
3 Φ , (12.2.18b)
∂c ∂c

where Q0 and QIN T are the free and interaction terms of Q, and O = 2 [c, ∂/∂c].
Each term in (12.2.18a) is separately BRST invariant. The BRST invariance of the
second term follows from the associativity property of the ∗ product. Due to the
−1/3 in (12.2.18b) one can easily show that all φ2 ψ terms drop out, due just to the c
dependence of Φ and the cyclicity of QIN T . Such terms would contain auxiliary fields
which drop out of the free action. We would like these fields to occur only in a way
which could be eliminated by field redefinition, corresponding to maintaining a gauge
invariance of the free action at the interacting level, so we could choose the gauge
where these fields vanish. Unfortunately, this is not the case in (12.2.18), so allowing
this abelian gauge invariance would require adding some additional cubic-interaction
gauge-fixing term, each term of which contains auxiliary-field factors, such that the
undesired auxiliary fields are eliminated from the action.

Whereas the δ-functionals used in the light-cone formalism correspond to a “flat”
geometry (see chapt. 10), with all curvature in the boundary (specifically, the split-
ting point) rather than the surface itself, those used in Witten’s covariant formalism
correspond to the geometry (with the 3 external legs amputated)

XX   C 

X X   C  
    XXX C 





(The boundaries of the strings are on top; the folds along the bottom don’t affect
the intrinsic geometry.) All the curvature is concentrated in the cusp at the bottom
(a small circle around it subtends an angle of 3π), with no intrinsic curvature in the
boundary (all parts of the boundary form angles of π with respect to the surface).
Each string is “folded” in the middle (π/2), and thus the vertex δ-functionals equate
the coordinates Z(σ) of one string with Z(π − σ) for the next string for σ ∈ [0, π/2]
(and therefore with Z(π − σ) for the previous string for σ ∈ [π/2, π]). These δ
functions are easily seen to define an associative product: Two successive ∗ products
produce a configuration like the one above, but with 4 strings, and associativity is
just the cyclicity of this 4-string object (see (12.2.10)). (However, vertex factors can
ruin this associativity because of divergences of the coincidence of two such factors
in the product of two vertices, as in the superstring: see below.)
Similar remarks apply to the corresponding product implied by the δ functionals
of (9.2.23) of the light-cone formalism, but there associativity is violated by an amount
which is fixed by the light-cone 4-point interaction vertex. This is due to the fact that
in the light-cone formalism there is a 4-point graph where a string has a string split
off from one side, followed by an incoming string joining onto the same side. If no
conformal transformation is made, this graph is nonplanar, unlike the graph where
this splitting and joining occur on opposite sides. There is a similar graph where
the splitting and joining occur on the opposite side from the first graph, and these
2 graphs are continuously related by a graph with a 4-point vertex as described in
chapt. 10. The σ-position of the interaction point in the surface of the string varies
from one end of the string to the other, with the vertex having this point at an end
being the same as the limit of the 1 of the other 2 graphs where the propagator has
vanishing length. On the other hand, in the covariant formalism the limits of the
12.2. Midpoint interaction 223

2 corresponding graphs are the same, so no 4-point vertex is needed to interpolate
between them.

The appropriate vertex factor for the midpoint interaction follows directly from
the quantum mechanics with bosonized ghosts. From (8.1.28a), since w = −1 for 
there is an 
C, d2 σ e−1 32 Rχ term which contributes only where there is curvature.
Thus, the total contribution of the curvature at the cusp to the path integral is an
extra factor of an exponential whose exponent is 3/2 times χ evaluated at the cusp.
The BRST invariance of the first-quantized action then guarantees that the vertex
conserves Q0 , and the coefficient of the curvature term compensates for the anomaly
in ghost-number conservation at the cusp. (Similar remarks apply for fermionized
ghosts using the Lorentz connection term.) Alternatively, the coefficient follows from
considering the ghost number of the fields and what ghost number is required for the
vertex to give the same matrix elements for physical polarizations as in the light-
cone formalism: In terms of bosonized-ghost coordinates, any physical state must be
∼ e−q̂/2 by (8.1.19,21a). Since the δ functional and functional integration in χ have
no such factors, and the vertex factor can be only at the cusp (otherwise it destroys
the above properties of the δ functionals), it must be e3χ/2 evaluated at the cusp to
cancel the q̂-dependence of the 3 fields. In terms of the original fermionic ghosts,
we use the latter argument, since the anomalous curvature term doesn’t show up
in the classical mechanics lagrangian (although a similar argument could be made
by considering quantum mechanical corrections). Then the physical states have no
dependence on c, while the vertex has a dc integration for each of the 3 coordinates,
and a single δ function for overall conservation of the “momentum” c. (A similar
argument follows from working in terms of Fourier transformed fields which depend
instead on the “coordinate” ∂/∂c.) The appropriate vertex factor is thus

C( π2 )C̃( π2 ) ∼ Ĉ( π2 )Ĉ(− π2 ) , (12.2.19a)

or, in terms of bosonized ghosts (but still for fields with fermionic ghost coordinates)

e2χ(π/2) ∼ eχ̂(π/2) eχ̂(−π/2) , (12.2.19b)

where π/2, the midpoint of each string, is the position of the cusp. (The difference in
the vertex factor for different coordinates is analogous to the fact that the “scalar”
(−g)−1/2 δ D (x − x ) in general relativity has different expressions for g in different
coordinate systems.) The vertex, including the factor (12.2.19), can be considered
a Heisenberg-picture vacuum in the same way as in the light-cone formalism, where

the vertex |V
= e∆ |0
in (9.2.1) was the effect of acting on the interaction-picture
vacuum with the S-matrix (of the first-quantized theory). However, in this case the
vacuum includes vertex factors because the appropriate vacuum is not the tachyon
one but rather the one left invariant by the Sp(2) subalgebra of the Virasoro algebra
[12.5] (see sect. 8.1). This is a consequence of the Sp(2) symmetry of the tree graphs.
Because of the midpoint form of the interaction there is a global symmetry cor-
responding to conformal transformations which leave the midpoint fixed. In second-
quantized notation, any operator which is the bracket of (the interacting) Q with
something itself has a vanishing bracket with Q, and is therefore simultaneously
BRST invariant and generates a global symmetry of the action (because Q is the
action). In particular, we can consider
δ δ

Q, †
Φ − Φ ∼ 
Φ† G(σ)  − σ) Φ ,
− G(π (12.2.20)

δ C(σ) 
δ C(π − σ) c

where the 2 δ/δ C terms cancel in the interaction term because of the form of the
overlap integral for the vertex (and the location of the vertex factor (12.2.19) at the
midpoint), and the surviving free term comes from the first-quantized expression for
G = {Q0 , δ/δ C}.
 Thus, this subalgebra of the Virasoro algebra remains a global
invariance at the interacting level (without becoming inhomogeneous in the fields).
The mode expansion of Witten’s vertex can be evaluated [12.6,7] as in the light-
cone case (sect. 9.2). (Partial evaluations were given and BRST invariance was also
studied in [12.8].) Now
1 δ 
∆= 2 ∆(P̌ , P̌ ) − ∆ Č , . (12.2.21)
δ Č
(C has weight w = 2.) The map from the ρ plane to the z plane can be found from
the following sequence of conformal transformations:

←2 x iπ/2 1, 3 → ρ = ln ζ

ζ =i = eρ

←3 2 1→
12.2. Midpoint interaction 225

1,3 • x 0 •2 η = λ3/2 =

• $
x 0 •2 λ= = η 2/3

3& %

• • •
3 2 1
(The bold-face numbers label the ends of the strings.) This maps the string from an
infinite rectangle (ρ) to the upper-half plane (ζ) to the interior of the unit circle (η) to
a different circle with all three strings appearing on the same sheet (λ) to the upper-
half plane with all strings on one sheet (z). If the cut for λ(η) is chosen appropriately
(the positive imaginary axis of the η plane), the cut under which the third string is
hidden is along the part of the imaginary ρ axis below iπ/2. (More conveniently, if
the cut is taken in the negative real direction in the η plane, then it’s in the positive
real direction in the ρ plane, with halves of 2 strings hidden under the cut.) Since the
last transformation is projective, we can drop it. (Projective transformations don’t
affect equations like (9.2.14).)
Unfortunately, although the calculation can still be performed [12.7], there is now
no simple analog to (9.2.18). It’s easier to use instead a map similar to (9.2.16) by
considering a 6-string δ functional with pairs of strings identified [12.6]: Specifically,
we replace the last 2 maps above with

• • 1$
i − ẑ
5• x 0 •2 λ̂ = = η 1/3
i + ẑ
• • 3%

1 − λ̂
ẑ = i
• • • • • 1 + λ̂
4 3 2 1 6 5→

and thus, relabeling r → r + 1 (or performing an equivalent rotation of the λ̂ circle)

ρ= p+r ln(ẑ − z r ) − ln 3 ,
√ 1 1 √
p+r = (−1)r+1 z r = ( 3, √ , 0, − √ , − 3, ∞) .
, (12.2.22)
3 3
We can then use the same procedure as the light cone. However, it turns out to be
more convenient to evaluate the contour integrals in terms of ζ rather than ẑ. Also,
instead of applying (9.2.18) to (12.2.22), we apply it to the corresponding expression
for λ̂:

ρ= αr ln(λ̂ − λr ) − 14 iπ ,
p+r = (−1)r , λr = e−iπ(r−2)/3 . (12.2.23)
Reexpressing (9.2.18) in terms of λ̂, we find
∂ ∂ 1 1
+  ln(λ̂ − λ̂ ) = 1
6 (λ̂3 + λ̂3 ) + (λ̂λ̂2 + λ̂2 λ̂ ) + + .
∂ρ ∂ρ λ̂λ̂2 λ̂2 λ̂
Using the conservation laws, the first set of terms can be dropped. Since it’s actually
λ = λ̂2 (or z), and not λ̂, for which the string is mapped to the complex plane
(λ̂ describes a 6-string vertex, and thus double counts), the ln we actually want to
evaluate is
ln(λ − λ ) = ln(λ̂ − λ̂ ) + ln(λ̂ + λ̂ ) . (12.2.25)
This just says that the general coefficients N rs in ∆ multiplying oscillators from
string r times those from string s is related to the corresponding fictitious 6-string
coefficients N rs
N rs = N  . (12.2.26)
rs r,s+3

The contour integrals can now be evaluated over ζ in terms of
 1/3 ∞
 2/3 ∞

1+x n 1+x
= an x , = bn xn , (12.2.27)
1−x 0 1−x 0
These coefficients satisfy the recursion relations

(n + 1)an+1 = 23 an + (n − 1)an−1 , (n + 1)bn+1 = 43 bn + (n − 1)bn−1 , (12.2.28)

which can be derived by appropriate manipulations of the corresponding contour
integrals: e.g.,
.  1/3
dx 1 1+x
an =
0 2πix xn 1−x
.  1/3 
dx 1 3 2 1+x
= 2 (1 − x ) . (12.2.29)
0 2πix xn 1−x
12.2. Midpoint interaction 227

Because of i’s relative to (12.2.27) appearing in the actual contour integrals, we use
instead the coefficients

(−1)n/2 (n even)
An = an · , (12.2.30)
(−1)(n−1)/2 (n odd)
and similarly for B n . We finally obtain an expression similar to (9.2.27), except that
we must use (12.2.26), and

N rsmn = m n M rsmn ,
+ p+s

M r,r+t,mn = 13 cmnt Am B n + (−1)m+n+t B m An ,

(−1)m Re(eit2π/3 ) (m + n even)
cmnt = . (12.2.31)
Im(eit2π/3 ) (m + n odd)
The terms for n = m = 0 or n = 0 = m can be evaluated by taking the appropriate
limit (n → m or n → 0). m = n = 0 can then be evaluated separately, using
(9.2.22b), (12.2.22), and (12.2.26). The final result is

 1 33 2
∆(Ψ̌1 , Ψ̌2 ) = − ψ 1 Nψ 2 − 4 ln p , (12.2.32)
where is over r, s = 1, 2, 3 and m, n = 0, 1, . . . , ∞ except for the term m = n = 0.
As for (9.2.27), ψ refers to all sets of oscillators, with ψ replaced with p+ 1−w ψ for
oscillators of weight w. In this case we use (12.2.21), and the p+ ’s are all ±1, so for

the ghosts there is an extra sign factor p+r p+s for N rsmn .

There are a number of problems to resolve for this formalism: (1) In calculating
S-matrix elements, the 4-point function is considerably more difficult to calculate than
in the light-cone formalism [12.9], and the conformal maps are so complicated that
it’s not yet known how to derive even the 5-point function for tachyons, although ar-
guments have been given for equivalence to the light-cone/external-field result [12.10].
(2) It doesn’t seem possible to derive an external-field approach to interactions, since
the string lengths are all fixed to be π. In the light-cone formalism the external-field
approach follows from choosing the Lorentz frame where all but 2 of the string lengths
(i.e., p+ ’s) vanish. (Thus, e.g., in the 3-string vertex 1 string reduces to a point on
the boundary, reducing to a vertex as in sect. 9.1.) This is related to the fact that I
of (12.2.11) is just the harmonic oscillator ground state at vanishing momentum (and
length) for the light-cone formalism, but for this formalism it’s ∼ δ[X(σ) − X(π − σ)].
(3) The fact that the gauge-invariant vertex is so different from the light-cone vertex
indicates that gauge-fixing to the light-cone gauge should be difficult. Furthermore,

the light-cone formalism requires a 4-point interaction in the action, whereas this co-
variant formalism doesn’t. Perhaps a formalism with a larger gauge invariance exists
such that these 2 formalisms are found by 2 different types of gauge choices. (4)
There is some difficulty in extending the discussion of sect. 11.1 for the closed string
to the interacting case, since the usual form of the physical-state vertex requires that
the vertex be related to the product of open-string vertices for the clockwise and
counterclockwise states, multiplied by certain vertex factors which don’t exist in this
formalism (although they would in a formalism more similar to the light-cone one,
since the light-cone formalism has more zero-mode conservation laws). This is partic-
ularly confusing since open strings generate closed ones at the 1-loop level. However,
some progress in understanding these closed strings has been made [12.11]. Also, a
general analysis has been made of some properties of the 3-point closed-string vertex
required by consistency of the 1-loop tadpole and 4-string tree graphs [12.12], using
techniques which are applicable to vertices more general than δ-functionals [12.13].
The gauge-fixing of this formalism with a BRST algebra that closes on shell has
been studied [12.14]. It has been shown both in the formalism of light-cone-like closed
string theory [12.15] and for the midpoint-interaction open string theory [12.16] that
the kinetic term can be obtained from an action with just the cubic term by giving
an appropriate vacuum value to the string field. However, whereas in the former case
(barring difficulties in loops mentioned above) this vacuum value is natural because
of the vacuum value of the covariant metric field for the graviton, in the latter case
there is no classical graviton in the open string theory, so the existence (or usefulness)
of such a mechanism is somewhat confusing.
The midpoint-interaction formulation of the open superstring (as a truncated
spinning string) has also been developed [11.3,12.17]. The supersymmetry algebra
closes only on shell, and the action apparently also needs (at least) 4-point interactions
to cancel divergences in 4-point amplitudes due to coincidence of vertex operator
factors (both of which occur at the midpoint) [12.18]. Such interactions might be of
the same type needed in the light-cone formulation (chapt. 10).

(1) Check the BRST invariance of (12.2.18).
(2) Find the transformation of ln(z−z  ) under the projective transformation z → az+b
(and similarly for z  ). Use the conservation law r ψ 0r = 0 to show that (9.2.14)
is unaffected.
Exercises 229

(3) Derive the last term of (12.2.32).



[0.1] M.B. Green, J.H. Schwarz, and E. Witten, Superstring theory, v. 1-2 (Cambridge
University, Cambridge, 1987).
[0.2] V. Gates, E. Kangaroo, M. Roachcock, and W.C. Gall, The super G-string, in Unified
string theories, eds. M. Green and D. Gross, Proc. of Santa Barbara Workshop, Jul.
29 - Aug. 16, 1985 (World Scientific, Singapore, 1986) p. 729;
Super G-string field theory, in Superstrings, Cosmology and Composite Structures,
eds. S.J. Gates, Jr. and R.N. Mohapatra, Proc. of International Workshop on Super-
strings, Composite Structures, and Cosmology, College Park, MD, March 11-18, 1987
(World Scientific, Singapore, 1987) p. 585.

Chapter 1

[1.1] P. van Nieuwenhuizen, Phys. Rep. 68 (1981) 189.
[1.2] M.T. Grisaru and W. Siegel, Nucl. Phys. B201 (1982) 292, B206 (1982) 496;
N. Marcus and A. Sagnotti, Phys. Lett. 135B (1984) 85;
M.H. Goroff and A. Sagnotti, Phys. Lett. 160B (1985) 81.
[1.3] Dual theory, ed. M. Jacob (North-Holland, Amsterdam, 1974);
J. Scherk, Rev. Mod. Phys. 47 (1975) 123;
P.H. Frampton, Dual resonance models and string theories (World Scientific, Singa-
pore, 1986).
[1.4] S. Mandelstam, Phys. Rep. 13 (1974) 259.
[1.5] J.H. Schwarz, Phys. Rep. 89 (1982) 223.
[1.6] G. ’t Hooft, Nucl. Phys. B72 (1974) 461.
[1.7] S. Weinberg, Critical phenomena for field theorists, in Understanding the fundamental
constituents of matter, proc. of the International School of Subnuclear Physics, Erice,
1976, ed. A. Zichichi (Plenum, New York, 1978) p. 1;
Ultraviolet divergences in quantum theories of gravitation, in General relativity: an
Einstein centenary survey, eds. S.W. Hawking and W. Israel (Cambridge University,
Cambridge, 1979) p. 790.
[1.8] G. ’t Hooft, Phys. Lett. 109B (1982) 474.
[1.9] T.H.R. Skyrme, Proc. Roy. Soc. A260 (1961) 227;
E. Witten, Nucl. Phys. B223 (1983) 433.
[1.10] D.J. Gross, A. Neveu, J. Scherk, and J.H. Schwarz, Phys. Lett. 31B (1970) 592.
[1.11] M.B. Green and J.H. Schwarz, Phys. Lett. 149B (1984) 117.
[1.12] M.B. Green and J.H.Schwarz, Phys. Lett. 109B (1982) 444.
[1.13] D.J. Gross, J.A. Harvey, E. Martinec, and R. Rohm, Phys. Rev. Lett. 54 (1985) 502,
Nucl. Phys. B256 (1985) 253, Nucl. Phys. B267 (1986) 75.
[1.14] L.J. Dixon and J.A. Harvey, Nucl. Phys. B274 (1986) 93;
L. Alvarez-Gaumé, P. Ginsparg, G. Moore, and C. Vafa, Phys. Lett. 171B (1986) 155;
H. Itoyama and T.R. Taylor, Phys. Lett. 186B (1987) 129.
[1.15] E. Witten, Phys. Lett. 149B (1984) 351.
[1.16] P. Candelas, G.T. Horowitz, A. Strominger, and E. Witten, Nucl. Phys. B258 (1985)

L. Dixon, J.A. Harvey, C. Vafa, and E. Witten, Nucl. Phys. B261 (1985) 678, 274
(1986) 285;
K.S. Narain, Phys. Lett. 169B (1986) 41;
H. Kawai, D.C. Lewellen, and S.-H.H. Tye, Nucl. Phys. B288 (1987) 1;
I. Antoniadis, C. Bachas, and C. Kounnas, Nucl. Phys. B289 (1987) 87;
R. Bluhm, L. Dolan, and P. Goddard, Nucl. Phys. B289 (1987) 364, Unitarity and mod-
ular invariance as constraints on four-dimensional superstrings, Rockefeller preprint
RU87/B1/25 DAMTP 88/9 (Jan. 1988).
[1.17] S.J. Gates, Jr., M.T. Grisaru, M. Roček, and W. Siegel, Superspace, or One thousand
and one lessons in supersymmetry (Benjamin/Cummings, Reading, 1983);
J. Wess and J. Bagger, Supersymmetry and supergravity (Princeton University, Prince-
ton, 1983);
R.N. Mohapatra, Unification and supersymmetry: the frontiers of quark-lepton physics
(Springer-Verlag, New York, 1986);
P. West, Introduction to supersymmetry and supergravity (World Scientific, Singa-
pore, 1986).

Chapter 2

[2.1] S. Weinberg, Phys. Rev. 150 (1966) 1313;
Kogut and D.E. Soper, Phys. Rev. D1 (1970) 2901.
[2.2] K . Bardakçi and M.B. Halpern, Phys. Rev. 176 (1968) 1686.
[2.3] W. Siegel and B. Zwiebach, Nucl. Phys. B282 (1987) 125.
[2.4] S.J. Gates, Jr., M.T. Grisaru, M. Roček, and W. Siegel, Superspace, or One thousand
and one lessons in supersymmetry (Benjamin/Cummings, Reading, 1983) p. 74.
[2.5] A. J. Bracken, Lett. Nuo. Cim. 2 (1971) 574;
A.J. Bracken and B. Jessup, J. Math. Phys. 23 (1982) 1925.
[2.6] W. Siegel, Nucl. Phys. B263 (1986) 93.
[2.7] W. Siegel, Phys. Lett. 142B (1984) 276.
[2.8] G. Parisi and N. Sourlas, Phys. Rev. Lett. 43 (1979) 744.
[2.9] R. Delbourgo and P.D. Jarvis, J. Phys. A15 (1982) 611;
J. Thierry-Mieg, Nucl. Phys. B261 (1985) 55;
J.A. Henderson and P.D. Jarvis, Class. and Quant. Grav. 3 (1986) L61.

Chapter 3

[3.1] P.A.M. Dirac, Proc. Roy. Soc. A246 (1958) 326;
L.D. Faddeev, Theo. Math. Phys. 1 (1969) 1.
[3.2] W. Siegel, Nucl. Phys. B238 (1984) 307.
[3.3] C. Becchi, A. Rouet, and R. Stora, Phys. Lett. 52B (1974) 344;
I.V. Tyutin, Gauge invariance in field theory and in statistical physics in the operator
formulation, Lebedev preprint FIAN No. 39 (1975), in Russian, unpublished;
T. Kugo and I. Ojima, Phys. Lett. 73B (1978) 459;
L. Baulieu, Phys. Rep. 129 (1985) 1.
[3.4] M. Kato and K. Ogawa, Nucl. Phys. B212 (1983) 443;

S. Hwang, Phys. Rev. D28 (1983) 2614;
K. Fujikawa, Phys. Rev. D25 (1982) 2584.
[3.5] N. Nakanishi, Prog. Theor. Phys. 35 (1966) 1111;
B. Lautrup, K. Dan. Vidensk. Selsk. Mat. Fys. Medd. 34 (1967) No. 11, 1.
[3.6] G. Curci and R. Ferrari, Nuo. Cim. 32A (1976) 151, Phys. Lett. 63B (1976) 91;
I. Ojima, Prog. Theo. Phys. 64 (1980) 625;
L. Baulieu and J. Thierry-Mieg, Nucl. Phys. B197 (1982) 477.
[3.7] L. Baulieu, W. Siegel, and B. Zwiebach, Nucl. Phys. B287 (1987) 93.
[3.8] S. Ferrara, O. Piguet, and M. Schweda, Nucl. Phys. B119 (1977) 493.
[3.9] J. Thierry-Mieg, J. Math. Phys. 21 (1980) 2834.
[3.10] E.S. Fradkin and G.A. Vilkovisky, Phys. Lett. 55B (1975) 224;
I.A. Batalin and G.A. Vilkovisky, Phys. Lett. 69B (1977) 309;
E.S. Fradkin and T.E. Fradkina, Phys. Lett. 72B (1978) 343;
M. Henneaux, Phys. Rep. 126 (1985) 1.
[3.11] M. Quirós, F.J. de Urries, J. Hoyos, M.L. Mazón, and E. Rodriguez, J. Math. Phys.
22 (1981) 1767;
L. Bonora and M. Tonin, Phys. Lett. 98B (1981) 48.
[3.12] S. Hwang, Nucl. Phys. B231 (1984) 386;
F.R. Ore, Jr. and P. van Nieuwenhuizen, Nucl. Phys. B204 (1982) 317.
[3.13] W. Siegel and B. Zwiebach, Nucl. Phys. B288 (1987) 332.
[3.14] W. Siegel and B. Zwiebach, Nucl. Phys. B299 (1988) 206.
[3.15] W. Siegel, Nucl. Phys. B284 (1987) 632.
[3.16] W. Siegel, Universal supersymmetry by adding 4+4 dimensions to the light cone,
Maryland preprint UMDEPP 88-231 (May 1988).

Chapter 4

[4.1] W. Siegel and B. Zwiebach, Nucl. Phys. B263 (1986) 105.
[4.2] T. Banks and M.E. Peskin, Nucl. Phys. B264 (1986) 513;
K. Itoh, T. Kugo, H. Kunitomo, and H. Ooguri, Prog. Theo. Phys. 75 (1986) 162.
[4.3] M. Fierz and W. Pauli, Proc. Roy. Soc. A173 (1939) 211;
S.J. Chang, Phys. Rev. 161 (1967) 1308;
L.P.S. Singh and C.R. Hagen, Phys. Rev. D9 (1974) 898;
C. Fronsdal, Phys. Rev. D18 (1978) 3624;
T. Curtright, Phys. Lett. 85B (1979) 219;
B. deWit and D.Z. Freedman, Phys. Rev. D21 (1980) 358;
T. Curtright and P.G.O. Freund, Nucl. Phys. B172 (1980) 413;
T. Curtright, Phys. Lett. 165B (1985) 304.
[4.4] W. Siegel, Phys. Lett. 149B (1984) 157, 151B (1985) 391.
[4.5] W. Siegel, Phys. Lett. 149B (1984) 162; 151B (1985) 396.
[4.6] E.S. Fradkin and V.I. Vilkovisky, Phys. Lett. 73B (1978) 209.
[4.7] W. Siegel and S.J. Gates, Jr., Nucl. Phys. B147 (1979) 77;
S.J. Gates, Jr., M.T. Grisaru, M. Roček, and W. Siegel, Superspace, or One thousand
and one lessons in supersymmetry (Benjamin/Cummings, Reading, 1983) p. 242.
[4.8] E. Witten, Nucl. Phys. B268 (1986) 253.
[4.9] A. Neveu, H. Nicolai, and P.C. West, Phys. Lett. 167B (1986) 307.
[4.10] W. Siegel and B. Zwiebach, Phys. Lett. 184B (1987) 325.

[4.11] M. Scheunert, W. Nahm, and V. Rittenberg, J. Math. Phys. 18 (1977) 155.
[4.12] A. Galperin, E. Ivanov, S. Kalitzin, V. Ogievetsky, and E. Sokatchev, JETP Lett. 40
(1984) 912, Class. Quant. Grav. 1 (1984) 469, 2 (1985) 601;
W. Siegel, Class. Quant. Grav. 2 (1985) 439;
E. Nissimov, S. Pacheva, and S. Solomon, Nucl. Phys. B296 (1988) 462, 297 (1988)
R.E. Kallosh and M.A. Rahmanov, Covariant quantization of the Green-Schwarz su-
perstring, Lebedev preprint (March 1988).
[4.13] J.P. Yamron, Phys. Lett. 174B (1986) 69.
[4.14] G.D. Daté, M. Günaydin, M. Pernici, K. Pilch, and P. van Nieuwenhuizen, Phys. Lett.
171B (1986) 182.
[4.15] H. Terao and S. Uehara, Phys. Lett. 173B (1986) 134;
T. Banks, M.E. Peskin, C.R. Preitschopf, D. Friedan, and E. Martinec, Nucl. Phys.
B274 (1986) 71;
A. LeClair and J. Distler, Nucl. Phys. B273 (1986) 552.

Chapter 5

[5.1] A. Barducci, R. Casalbuoni, and L. Lusanna, Nuo. Cim. 35A (1976) 377;
L. Brink, S. Deser, B. Zumino, P. DiVecchia, and P. Howe, Phys. Lett. 64B (1976)
P.A. Collins and R.W. Tucker, Nucl. Phys. B121 (1977) 307.
[5.2] S. Deser and B. Zumino, Phys. Lett. 62B (1976) 335.
[5.3] E.C.G. Stueckelberg, Helv. Phys. Acta 15 (1942) 23;
R.P. Feynman, Phys. Rev. 80 (1950) 440.
[5.4] J. Polchinski, Commun. Math. Phys. 104 (1986) 37.
[5.5] J. Schwinger, Phys. Rev. 82 (1951) 664.
[5.6] R.J.Eden, P.V. Landshoff, D.I. Olive, and J.C. Polkinghorne, The analytic S-matrix
(Cambridge University, Cambridge, 1966) p. 152.
[5.7] ibid., p. 88;
J.D. Bjorken and S.D. Drell, Relativistic quantum fields (McGraw-Hill, New York,
1965) p. 225.
[5.8] J. Iliopoulos, G. Itzykson, and A. Martin, Rev. Mod. Phys. 47 (1975) 165.
[5.9] M.B. Halpern and W. Siegel, Phys. Rev. D16 (1977) 2486.
[5.10] F.A. Berezin and M.S. Marinov, Ann. Phys. (NY) 104 (1977) 336;
A. Barducci, R. Casalbuoni, and L. Lusanna, Nucl. Phys. B124 (1977) 93;
A.P. Balachandran, P. Salomonson, B.-S. Skagerstam, and J.-O. Winnberg, Phys. Rev.
D15 (1977) 2308.
[5.11] H. Georgi, Lie algebras in particle physics: from isospin to unified theories (Ben-
jamin/Cummings, Reading,1982) p. 192.
[5.12] S.J. Gates, Jr., M.T. Grisaru, M. Roček, and W. Siegel, Superspace, or One thousand
and one lessons in supersymmetry (Benjamin/Cummings, Reading, 1983) pp. 70, 88.
[5.13] L. Brink, J.H. Schwarz, and J. Scherk, Nucl. Phys. B121 (1977) 77;
W. Siegel, Phys. Lett. 80B (1979) 220.
[5.14] M.T. Grisaru, W. Siegel, and M. Roček, Nucl. Phys. B159 (1979) 429;
M.T. Grisaru and W. Siegel, Nucl. Phys. B201 (1982) 292, B206 (1982) 496;
S.J. Gates, Jr., M.T. Grisaru, M. Roček, and W. Siegel, Superspace, pp. 25, 382, 383.

[5.15] W. Siegel, Nucl. Phys. B156 (1979) 135.
[5.16] J. Koller, Nucl. Phys. B222 (1983) 319;
P.S. Howe, G. Sierra, and P.K. Townsend, Nucl. Phys. B221 (1983) 331;
J.P. Yamron and W. Siegel, Nucl. Phys. B263 (1986) 70.
[5.17] J. Wess and B. Zumino, Nucl. Phys. B70 (1974) 39;
R. Haag, J.T. L opuszański, and M. Sohnius, Nucl. Phys. B88 (1975) 257;
W. Nahm, Nucl. Phys. B135 (1978) 149.
[5.18] P. Ramond, Physica 15D (1985) 25.
[5.19] W. Siegel, Free field equations for everything, in Strings, Cosmology, Composite Struc-
tures, March 11-18, 1987, College Park, Maryland, eds. S.J. Gates, Jr. and R.N. Mo-
hapatra (World Scientific, Singapore, 1987).
[5.20] A. Salam and J. Strathdee, Nucl. Phys. B76 (1974) 477;
S.J. Gates, Jr., M.T. Grisaru, M. Roček, and W. Siegel, Superspace, pp. 72-73;
R. Finkelstein and M. Villasante, J. Math. Phys. 27 (1986) 1595.
[5.21] W. Siegel, Phys. Lett. 203B (1988) 79.
[5.22] J. Strathdee, Inter. J. Mod. Phys. A2 (1987) 273.
[5.23] W. Siegel, Class. Quantum Grav. 2 (1985) L95.
[5.24] W. Siegel, Phys. Lett. 128B (1983) 397.
[5.25] W. Siegel and S.J. Gates, Jr., Nucl. Phys. B189 (1981) 295.
[5.26] S. Mandelstam, Nucl. Phys. B213 (1983) 149.
[5.27] L. Brink, O. Lindgren, and B.E.W. Nilsson, Nucl. Phys. B212 (1983) 401.
[5.28] R. Casalbuoni, Phys. Lett. 62B (1976) 49;
L. Brink and J.H. Schwarz, Phys. Lett. 100B (1981) 310.
[5.29] A.R. Miković and W. Siegel, On-shell equivalence of superstrings, Maryland preprint
88-218 (May 1988).
[5.30] K. Kamimura and M. Tatewaki, Phys. Lett. 205B (1988) 257.
[5.31] R. Penrose, J. Math. Phys. 8 (1967) 345, Int. J. Theor. Phys. 1 (1968) 61 ;
M.A.H. MacCallum and R. Penrose, Phys. Rep. 6C (1973) 241.
[5.32] A. Ferber, Nucl. Phys. B132 (1978) 55.
[5.33] T. Kugo and P. Townsend, Nucl. Phys. B221 (1983) 357;
A. Sudbery, J. Phys. A17 (1984) 939;
K.-W. Chung and A. Sudbery, Phys. Lett. 198B (1987) 161.
[5.34] Z. Hasiewicz and J. Lukierski, Phys. Lett. 145B (1984) 65;
I. Bengtsson and M. Cederwall, Particles, twistors, and the division algebras, Ecole
Normale Supérieure preprint LPTENS-87/20 (May 1987).
[5.35] D.B. Fairlie and C.A. Manogue, Phys. Rev. D34 (1986) 1832, 36 (1987) 475;
J.M. Evans, Nucl. Phys. B298 (1988) 92.

Chapter 6

[6.1] P.A. Collins and R.W. Tucker, Phys. Lett. 64B (1976) 207;
L. Brink, P. Di Vecchia, and P. Howe, Phys. Lett. 65B (1976) 471;
S. Deser and B. Zumino, Phys. Lett. 65B (1976) 369.
[6.2] Y. Nambu, lectures at Copenhagen Symposium, 1970 (unpublished);
O. Hara, Prog. Theo. Phys. 46 (1971) 1549;
T. Goto, Prog. Theo. Phys. 46 (1971) 1560.
[6.3] S.J. Gates, Jr., R. Brooks, and F. Muhammad, Phys. Lett. 194B (1987) 35;

C. Imbimbo and A. Schwimmer, Phys. Lett. 193B (1987) 455;
J.M.F. Labastida and M. Pernici, Nucl. Phys. B297 (1988) 557;
L. Mezincescu and R.I. Nepomechie, Critical dimensions for chiral bosons, Miami
preprint UMTG-140 (Aug. 1987);
Y. Frishman and J. Sonnenschein, Gauging of chiral bosonized actions, Weizmann
preprint WIS-87/65/Sept.-PH (Sept. 1987);
R. Floreanini and R. Jackiw, Phys. Rev. Lett. 59 (1987) 1873;
S. Bellucci, R. Brooks, and J. Sonnenschein, Supersymmetric chiral bosons, SLAC
preprint SLAC-PUB-4458 MIT-CTP-1548 UCD-87-35 (Oct. 1987);
S.J. Gates, Jr. and W. Siegel, Leftons, rightons, nonlinear σ-models, and superstrings,
Maryland preprint UMDEPP 88-113 (Nov. 1987);
M. Henneaux and C. Teitelboim, Dynamics of chiral (self-dual) p-forms, Austin preprint
(Dec. 1987);
M. Bernstein and J. Sonnenschein, A comment on the quantization of chiral bosons,
SLAC preprint SLAC-PUB-4523 (Jan. 1988);
L. Faddeev and R. Jackiw, Hamiltonian reduction of unconstrained and constrained
systems, MIT preprint CTP#1567 (Feb. 1988);
J. Sonnenschein, Chiral bosons, SLAC preprint SLAC-PUB-4570 (March 1988).
[6.4] C.M. Hull, Covariant quantization of chiral bosons and anomaly cancellation, London
preprint Imperial/TP/87/88/9 (March 1988).
[6.5] P. Goddard, J. Goldstone, C. Rebbi, and C.B. Thorn, Nucl. Phys. B56 (1973) 109.

Chapter 7

[7.1] P. Ramond, Phys. Rev. D3 (1971) 86;
A. Neveu and J.H. Schwarz, Nucl. Phys. B31 (1971) 86, Phys. Rev. D4 (1971) 1109.
[7.2] D.B. Fairlie and D. Martin, Nuo. Cim. 18A (1973) 373, 21A (1974) 647;
L. Brink and J.-O. Winnberg, Nucl. Phys. B103 (1976) 445.
[7.3] F. Gliozzi, J. Scherk, and D.I. Olive, Phys. Lett. 65B (1976) 282, Nucl. Phys. B122
(1977) 253.
[7.4] Y. Iwasaki and K. Kikkawa, Phys. Rev. D8 (1973) 440.
[7.5] W. Siegel, Covariant approach to superstrings, in Symposium on anomalies, geometry
and topology, eds. W.A. Bardeen and A.R. White (World Scientific, Singapore, 1985)
p. 348.
[7.6] M.B. Green and J.H. Schwarz, Phys. Lett. 136B (1984) 367, Nucl. Phys. B243 (1984)

Chapter 8

[8.1] M. Virasoro, Phys. Rev. D1 (1970) 2933;
I.M. Gelfand and D.B. Fuchs, Functs. Anal. Prilozhen 2 (1968) 92.
[8.2] E. Del Giudice, P. Di Vecchia, and S. Fubini, Ann. Phys. 70 (1972) 378.
[8.3] F. Gliozzi, Nuovo Cim. Lett. 2 (1969) 846.
[8.4] E. Witten, Some remarks about string field theory, in Marstrand Nobel Sympos. 1986,
p. 70;
I.B. Frenkel, H. Garland, and G.J. Zuckerman, Proc. Natl. Acad. Sci. USA 83 (1986)

[8.5] P. Jordan, Z. Physik 93 (1935) 464;
M. Born and N.S. Nagendra Nath, Proc. Ind. Acad. Sci. 3 (1936) 318;
A. Sokolow, Phys. Z. der Sowj. 12 (1937) 148;
S. Tomonaga, Prog. Theo. Phys. 5 (1950) 544;
T.H.R. Skyrme, Proc. Roy. Soc. A262 (1961) 237;
D. Mattis and E. Lieb, J. Math. Phys. 6 (1965) 304;
B. Klaiber, in Lectures in theoretical physics, eds. A.O. Barut and W.E. Brittin (Gor-
don and Breach, New York, 1968) v. X-A, p. 141;
R.F. Streater and I.F. Wilde, Nucl. Phys. B24 (1970) 561;
J. Lowenstein and J. Swieca, Ann. Phys. (NY) 68 (1971) 172;
K. Bardakçi and M.B. Halpern, Phys. Rev. D3 (1971) 2493;
G. Dell’Antonio, Y. Frishman, and D. Zwanziger, Phys. Rev. D6 (1972) 988;
A. Casher, J. Kogut, and L. Susskind, Phys. Rev. Lett. 31 (1973) 792, Phys. Rev. D10
(1974) 732;
A. Luther and I. Peschel, Phys. Rev. B9 (1974) 2911;
A. Luther and V. Emery, Phys. Rev. Lett. 33 (1974) 598;
S. Coleman, Phys. Rev. D11 (1975) 2088;
B. Schroer, Phys. Rep. 23 (1976) 314;
S. Mandelstam, Phys. Rev. D11 (1975) 3026;
J. Kogut and L. Susskind, Phys. Rev. D11 (1975) 3594.
[8.6] M.B. Halpern, Phys. Rev. D12 (1975) 1684;
I.B. Frenkel and V.G. Kač, Inv. Math. 62 (1980) 23;
I.B. Frenkel, J. Func. Anal. 44 (1981) 259;
G. Segal, Comm. Math. Phys. 80 (1981) 301;
P. Goddard and D. Olive, Vertex operators in mathematics and physics, eds. J. Lep-
owsky, S. Mandelstam, and I.M. Singer (Springer-Verlag, New York, 1985) p. 51.
[8.7] R. Marnelius, Nucl. Phys. B211 (1983) 14.
[8.8] W. Siegel, Phys. Lett. 134B (1984) 318.
[8.9] A.M. Polyakov, Phys. Lett. 103B (1981) 207.
[8.10] A.A. Tseytlin, Phys. Lett. 168B (1986) 63;
S.R. Das and M.A. Rubin, Phys. Lett. 169B (1986) 182, Prog. Theor. Phys. Suppl. 86
(1986) 143, Phys. Lett. 181B (1986) 81;
T. Banks, D. Nemeschansky, and A. Sen, Nucl. Phys. B277 (1986) 67;
G.T. Horowitz and A. Strominger, Phys. Rev. Lett. 57 (1986) 519;
G. Münster, Geometric string field theory, DESY preprint DESY 86-045 (Apr. 1986).
[8.11] H. Aoyama, Nucl. Phys. B299 (1988) 379.
[8.12] D.Z. Freedman, J.I. Latorre, and K. Pilch, Global aspects of the harmonic gauge in
bosonic string theory, MIT preprint CTP#1559 (Jan. 1988).

Chapter 9

[9.1] K. Bardakçi and H. Ruegg, Phys. Rev. 181 (1969) 1884;
C.J. Goebel and B. Sakita, Phys. Rev. Lett. 22 (1969) 257;
Chan H.-M. and T.S. Tsun, Phys. Lett. 28B (1969) 485;
Z. Koba and H.B. Nielsen, Nucl. Phys. B10 (1969) 633, 12 (1969) 517.
[9.2] G. Veneziano, Nuo. Cim. 57A (1968) 190.

[9.3] C. Lovelace, Phys. Lett. 135B (1984) 75;
D. Friedan, Z. Qiu, and S. Shenker, proc. APS Div. of Particles and Fields Conf.,
Santa Fe, eds. T. Goldman and M. Nieto (World Scientific, Singapore, 1984);
E. Fradkin and A. Tseytlin, Phys. Lett. 158B (1985) 316;
A. Sen, Phys. Rev. D32 (1985) 2102, Phys. Rev. Lett. 55 (1985) 1846;
C.G. Callan, E.J. Martinec, M.J. Perry, and D. Friedan, Nucl. Phys. B262 (1985) 593.
[9.4] S. Mandelstam, Nucl. Phys. B64 (1973) 205;
J.L. Torres-Hernández, Phys. Rev. D11 (1975) 3565;
H. Arfaei, Nucl. Phys. B85 (1975) 535, 112 (1976) 256.
[9.5] E. Cremmer and J.-L. Gervais, Nucl. Phys. B90 (1975) 410.
[9.6] W. Siegel, Nucl. Phys. B109 (1976) 244.
[9.7] M.B. Green and J.H. Schwarz, Phys. Lett. 151B (1985) 21.
[9.8] P.H. Frampton, P. Moxhay, and Y.J. Ng, Phys. Rev. Lett. 55 (1985) 2107, Nucl. Phys.
B276 (1986) 599;
L. Clavelli, Phys. Rev. D33 (1986) 1098, 34 (1986) 3262, Prog. Theor. Phys. Suppl.
86 (1986) 135.
[9.9] O. Alvarez, Nucl. Phys. B216 (1983) 125;
E. Martinec, Phys. Rev. D28 (1983) 2604;
G. Moore and P. Nelson, Nucl. Phys. B266 (1986) 58;
A. Cohen, G. Moore, P. Nelson, and J. Polchinski, Nucl. Phys. B267 (1986) 143;
S. Mandelstam, The interacting-string picture and functional integration, in Unified
string theories, eds. M. Green and D. Gross (World Scientific, Singapore, 1986) pp.
E. D’Hoker and D.H. Phong, Nucl. Phys. B269 (1986) 205;
M.A. Namazie and S. Rajeev, Nucl. Phys. B277 (1986) 332;
D. Friedan, E. Martinec, and S. Shenker, Nucl. Phys. B271 (1986) 93;
E. D’Hoker and D.H. Phong, Commun. Math. Phys. 104 (1986) 537, Phys. Rev. Lett.
56 (1986) 912;
E. Gava, R. Jengo, T. Jayaraman, and R. Ramachandran, Phys. Lett. 168B (1986)
G. Moore, P. Nelson, and J. Polchinski, Phys. Lett. 169B (1986) 47;
R. Catenacci, M. Cornalba, M. Martellini, C. Reina, Phys. Lett. 172B (1986) 328;
E. Martinec, Phys. Lett. 171B (1986) 189;
E. D’Hoker and D.H. Phong, Nucl. Phys. B278 (1986) 225;
S.B. Giddings and S.A. Wolpert, Comm. Math. Phys. 109 (1987) 177;
E. D’Hoker and S.B. Giddings, Nucl. Phys. B291 (1987) 90.
[9.10] E. D’Hoker and D.H. Phong, The geometry of string perturbation theory, Princeton
preprint PUPT-1039 (Feb. 1988), to appear in Rev. Mod. Phys. 60 (1988).
[9.11] J. Shapiro, Phys. Rev. D5 (1972) 1945;
E. Witten, Global anomalies in string theory, in Symposium on anomalies, geometry
and topology, eds. W.A. Bardeen and A.R. White (World Scientific, Singapore, 1985)
p. 348.

Chapter 10

[10.1] M. Kaku and K. Kikkawa, Phys. Rev. D10 (1974) 1110, 1823, M. Kaku, Phys. Rev.
D10 (1974) 3943;

E. Cremmer and J.-L. Gervais, Nucl. Phys. B76 (1974) 209.
[10.2] M.B. Green and J.H. Schwarz, Nucl. Phys. B243 (1984) 475.
[10.3] Chan H.M. and J. Paton, Nucl. Phys. B10 (1969) 519;
N. Marcus and A. Sagnotti, Phys. Lett. 119B (1982) 97.
[10.4] N. Berkovits, Nucl. Phys. B276 (1986) 650; Supersheet functional integration and
the interacting Neveu-Schwarz string, Berkeley preprint UCB-PTH-87/31 LBL-23727
(July 1987);
S.-J. Sin, Geometry of super lightcone diagrams and Lorentz invariance of lightcone
string field theory, (II) closed Neveu-Schwarz string, Berkeley preprint LBL-25120
UCB-PTH-88/09 (April 1988).
[10.5] S. Mandelstam, Nucl. Phys. B83 (1974) 413.
[10.6] T. Kugo, Prog. Theor. Phys. 78 (1987) 690;
S. Uehara, Phys. Lett. 196B (1987) 47;
S.-J. Sin, Lorentz invariance of light cone string field theories, Berkeley preprint LBL-
23715 (June 1987).
[10.7] N. Linden, Nucl. Phys. B286 (1987) 429.
[10.8] D.J. Gross and V. Periwal, Nucl. Phys. B287 (1987) 1.
[10.9] J. Greensite and F.R. Klinkhamer, Nucl. Phys. B281 (1987) 269, 291 (1987) 557;
Superstring amplitudes and contact interactions, Berkeley preprint LBL-23830 NBI-
HE-87-58 (Aug. 1987).
[10.10] M. Saadi, Intermediate closed strings as open strings going backwards in time, MIT
preprint CTP#1575 (April 1988).

Chapter 11

[11.1] C. Marshall and P. Ramond, Nucl. Phys. B85 (1975) 375;
A. Neveu, H. Nicolai, and P.C. West, Nucl. Phys. B264 (1986) 573;
A. Neveu, J.H. Schwarz, and P.C. West, Phys. Lett. 164B (1985) 51,
M. Kaku, Phys. Lett. 162B (1985) 97;
S. Raby, R. Slansky, and G. West, Toward a covariant string field theory, in Lewes
string theory workshop, eds. L. Clavelli and A. Halprin (World Scientific, Singapore,
[11.2] T. Banks and M.E. Peskin, Gauge invariant functional field theory for bosonic strings,
in Symposium on anomalies, geometry, topology, eds. W.A. Bardeen and A.R. White,
Mar. 28-30, 1985 (World Scientific, Singapore, 1985);
M. Kaku, Nucl. Phys. B267 (1986) 125;
D. Friedan, Nucl. Phys. B271 (1986) 540;
A. Neveu and P.C. West, Nucl. Phys. B268 (1986) 125;
K. Bardakçi, Nucl. Phys. B271 (1986) 561;
J.-L. Gervais, Nucl. Phys. B276 (1986) 339.
[11.3] E. Witten, Nucl. Phys. B276 (1986) 291;
T. Kugo and H. Terao, New gauge symmetries in Witten’s Ramond string field theory,
Kyoto preprint KUNS 911 HE(TH)88/01 (Feb. 1988).

Chapter 12

[12.1] A. Strominger, Phys. Rev. Lett. 58 (1987) 629.
[12.2] C.-Y. Ye, Covariantized closed string fields from light-cone, MIT preprint CTP#1566
(Feb. 1988).
[12.3] H. Hata, K. Itoh, T. Kugo, H. Kunitomo, and K. Ogawa, Phys. Lett. 172B (1986)
186, 195; Nucl. Phys. B283 (1987) 433; Phys. Rev. D34 (1986) 2360, 35 (1987) 1318,
1356; Prog. Theor. Phys. 77 (1987) 443, 78 (1987) 453.
[12.4] M.A. Awada, Phys. Lett. 172B (1986) 32;
A. Neveu and P. West, Phys. Lett. 168B (1986) 192.
[12.5] A. Hosoya and H. Itoyama, The vertex as a Bogoliubov transformed vacuum state in
string field theory, Fermilab preprint FERMILAB-PUB-87/111-T OUHET-12 (Jan.
T. Kugo, H. Kunitomo, and K. Suehiro, Prog. Theor. Phys. 78 (1987) 923.
[12.6] D.J. Gross and A. Jevicki, Nucl. Phys. B283 (1987) 1, 287 (1987) 225;
Z. Hlousek and A. Jevicki, Nucl. Phys. B288 (1987) 131.
[12.7] E. Cremmer, C.B. Thorn, and A. Schwimmer, Phys. Lett. 179B (1986) 57;
C.B. Thorn, The oscillator representation of Witten’s three open string vertex func-
tion, in Proc. XXIII Int. Conf. on High Energy Physics, July 16-23, 1986, Berkeley,
CA, ed. S.C. Loken (World Scientific, Singapore, 1987) v.1, p. 374.
[12.8] S. Samuel, Phys. Lett. 181B (1986) 255;
N. Ohta, Phys. Rev. D34 (1986) 3785;
K. Itoh, K. Ogawa, and K. Suehiro, Nucl. Phys. B289 (1987) 127;
A. Eastaugh and J.G. McCarthy, Nucl. Phys. B294 (1987) 845.
[12.9] S.B. Giddings, Nucl. Phys. B278 (1986) 242.
[12.10] S.B. Giddings and E. Martinec, Nucl. Phys. B278 (1986) 91;
S.B. Giddings, E. Martinec, and E. Witten, Phys. Lett. 176B (1986) 362.
[12.11] J.A. Shapiro and C.B. Thorn, Phys. Rev. D36 (1987) 432, Phys. Lett. 194B (1987)
D.Z. Freedman, S.B. Giddings, J.A. Shapiro, and C.B. Thorn, Nucl. Phys. B298 (1988)
[12.12] B. Zwiebach, Constraints on covariant theories for closed string fields, MIT preprint
CTP#1583 (April 1988); A note on covariant Feynman rules for closed strings, MIT
preprint CTP#1598 (May 1988).
[12.13] A. LeClair, M.E. Peskin, and C.R. Preitschopf, String field theory on the conformal
plane, I. Kinematical principles, II. Generalized gluing, SLAC preprints SLAC-PUB-
4306, 4307 (Jan. 1988), C.R. Preitschopf, The gluing theorem in the operator formu-
lation of string field theory, Maryland preprint UMDEPP-88-087 (Oct. 1987);
A. Eskin, Conformal transformations and string field redefinitions, MIT preprint
CTP#1560 (Feb. 1988).
[12.14] M. Bochicchio, Phys. Lett. 193B (1987) 31, 188B (1987) 330;
C.B. Thorn, Nucl. Phys. B287 (1987) 61.
[12.15] H. Hata, K. Itoh, T. Kugo, H. Kunitomo, and K. Ogawa, Phys. Lett. 175B (1986)
[12.16] G.T. Horowitz, J. Lykken, R. Rohm, and A. Strominger, Phys. Rev. Lett. 57 (1986)
[12.17] D.J. Gross and A. Jevicki, Nucl. Phys. B293 (1987) 29;

S. Samuel, Nucl. Phys. B296 (1988) 187.
[12.18] M.B. Green and N. Seiberg, Nucl. Phys. B299 (1988) 559;
C. Wendt, Scattering amplitudes and contact interactions in Witten’s superstring field
theory, SLAC preprint SLAC-PUB-4442 (Nov. 1987).


Action 10, 11.1, 11.2, 12.2
see BRST; Light cone; Yang-Mills Coherent states 5.5, 9.1
Anomalies Compactification 1.2, 1.3
see Graphs, anomalies Component expansions 11.2
Auxiliary fields 2.1, 4.1 Conformal transformations
D=2 6.2, 8.1, 9.1, 9.2
Bosonic strings 6.1, 7.1, 8.1 D>2 2.2, 2.3
Bosonization 8.1 Sp(2) 8.1, 9.2, 9.3, 12.2
Bound states 9.3 superconformal 5.4
Boundary conditions 6.1, 6.3 Conjugation, charge, complex, or hermitian
Bracket, covariant 3.4, 4.2 see Proper time reversal
BRST 3.1, 4.1, 11.1, 12.1 Constraints
action, gauge-invariant 4.1, 11.1, 12.2 see BRST
BRST1 3.2, 3.4, 4.1 Coset space methods 2.2,5.4
BRST2 3.2, 3.3 Covariantized light cone
closed strings 11.1 see Light cone, covariantized
extra modes 3.2, 4.3 Critical dimension 1.2, 7.1, 8.1
gauge fixing 3.2, 4.4, 5.5, 12.2
GL(1|1) 4.2, 8.2 DDF operators 8.1
GL(2|2) 3.6, 4.1, 8.3 Dimension, spacetime
IGL(1) 3.2, 4.2, 8.1, 12.1, 12.2 D=1 1.1, 2.6, 5.1
infinite tower of ghosts 5.5 D=2 1.1, 1.3, 3.1, 6.1
interacting 2.4, 3.2, 3.3, 3.4, 4.2, D=3 5.4, 7.3
12.1, 12.2 D=4 1.1, 1.3, 5.4, 7.3
IOSp(1,1|2) 3.3, 11.1 D=6 1.3, 5.4, 7.3
IOSp(2,2|4) 3.6, 4.1 D = 10 1.2, 1.3, 5.4, 7.3
IOSp(D,2|2) 2.6, 3.4 D = 26 1.2, 7.1, 8.1
IOSp(D+1,3|4) 3.6, 4.1, 5.5, 8.3 Dimensional reduction 2.3, 5.4
Lorentz gauge 3.2 Divergences
Lorentz gauge, string 8.3 see Graphs, loops
open strings 8.1, 8.3, 12.1, 12.2 Division algebras 5.4
OSp(1,1|2) 3.3, 3.4, 3.5, 3.6, 4.1, 4.5, Duality 9.1
8.2, 8.3
OSp(D,2|2) 2.6, 3.4 External fields
particles 5.2, 5.5 see Graphs, external fields
supersymmetry 5.5, 7.3
temporal gauge 3.1, 3.2 Fermionization 8.1
U(1,1|1,1) 3.6, 4.1, 4.5, 5.5 Fermions 3.5, 4.5, 5.3
Field, string 10.1, 11.1, 12.1
Canonical quantization Field equations
covariant 3.4, 4.2 field strengths 2.2, 5.4, 7.3
light cone 2.4 gauge fields
Chan-Paton factors see BRST; Yang-Mills
see Group theory indices Finiteness
Chiral scalars 3.1, 6.2 see Loops
Classical mechanics 4+4-extension
conformal gauge 6.2 see BRST: GL(2|2), IOSp(2,2|4), U(1,1|1,1)
gauge covariant 6.1 Functional integrals 9.2
light cone 6.3
particle 5.1 γ matrices 2.1, 3.5, 5.3, 5.4, 7.2, 7.3
Closed strings 1.1, 1.2, 6.1, 6.3, 7.1, 8.2, Gauge

see BRST; Light cone D=2 6.2, 8.3
Gauge fixing D>2
see BRST; Light cone see Gravity
Ghosts flat-space 2.1, 6.2
see BRST
Graphs Neveu-Schwarz-Ramond model
anomalies 1.2, 8.1, 9.3 see Spinning strings
covariantized light cone 2.6 No-ghost theorem 4.4, 4.5
external fields 9.1
light cone 2.5 Octonions 5.4
loops 1.1, 1.2, 9.3 Orientation
trees 9.2 see Twists
Gravity 4.1, 11.2 OSp
super 1.1 see BRST; Light cone, covariantized
Group theory indices 7.1, 9.2, 10
Gupta-Bleuler 3.1, 3.2, 5.4, 7.3, 8.1 Particle 1.1, 5.1
bosonic 5.1, 5.2
Hadrons 1.1 isospinning 5.3
Hamiltonian quantization 3.1 spinning 5.3
Heterotic strings supersymmetric 5.4
see Superstrings Path integrals 9.2
Perturbation theory
IGL(1) see Graphs
see BRST Poincaré algebra
Interacting string picture see Light cone, Poincaré algebra
see Graphs Polyakov approach
Interactions see Graphs
see BRST; Light cone Projective group 8.1, 9.3
IOSp Proper time τ
see BRST; Light cone, covariantized see Dimension, spacetime, D = 1
Proper time reversal 5.1, 6.3, 8.1, 8.3, 10
Klein transformation 7.2
Koba-Nielsen amplitude 9.1 QCD 1.1
Quantum mechanics
Lagrange multipliers 3.1 see BRST; Canonical quantization;
Length, string 6.3 Light cone; Path integrals
Light cone 2.1, 7.1, 10 Quaternions 5.4
actions 2.1, 5.5, 10
bosonic particles 5.1 Ramond-Neveu-Schwarz model
bosonic strings 6.3, 7.1, 10 see Spinning strings
covariantized 2.6, 10 Regge theory 1.1, 9.1
graphs 2.5, 9.2 Reparametrization invariance
interactions 2.1, 2.4, 9.2, 10 see Metric, D = 1 and D = 2
Poincaré algebra 2.3, 7.1, 7.2, 10
spinning strings 7.2 Scale invariance, 2D local (Weyl)
spinors 5.3 see Metric, D = 2
superparticles 5.4, 5.5 σ reversal
superstrings 7.3 see Twists
Yang-Mills 2.1, 2.4 S-matrix
Loops see Graphs
see Graphs, loops Spectrum, mass 7.1
Spinning strings 7.2
Metric Spinors
D=1 5.1 see Fermions; γ matrices; Supersymmetry

see the whole book
Stueckelberg fields 4.1
Superconformal transformations 5.4, 7.2
see Gravity, super
Superspin 5.4, 5.5
Superstrings (I, IIAB, heterotic) 1.2, 7.3, 9.3,
D=1 5.3
D=2 7.2
D>2 1.2, 1.3, 5.4, 5.5, 7.3

Tension, string (α ) 6.1
see Graphs, trees
Twistors 5.4
Twists 7.1, 10

Veneziano model
see Bosonic string
see BRST; Light cone
Virasoro operators
see Conformal transformations, 2D

World sheet
see Dimension, spacetime, D=2

Yang-Mills 2.1, 2.4, 3.1, 3.2, 3.3, 3.4, 4.1,
4.2, 4.4, 4.6, 11.2
super 5.4, 9.1